spectroscopic observations of the candidate sgb[e]/x-ray ...2 r. i. hynes et al.: spectroscopic...

Astronomy & Astrophysics manuscript no.(will be inserted by hand later)

Spectroscopic observations of thecandidate sgB[e]/X-ray binary CI Cam

R. I. Hynes1, J. S. Clark2, E. A. Barsukova3, P. J. Callanan4, P. A. Charles1, A. Collier Cameron5,S. N. Fabrika3, M. R. Garcia6, C. A. Haswell7,?, Keith Horne5, A. Miroshnichenko8,9, I. Negueruela10,

P. Reig11,12, W. F. Welsh13,?, and D. K. Witherick2

1 Department of Physics and Astronomy, University of Southampton, Southampton, SO17 1BJ, UK2 Department of Physics and Astronomy, University College London, Gower Street, London, WC1E 6BT, UK3 Special Astrophysical Observatory, Nizhnij Arkhyz, 369167, Russia4 University College, Department of Physics, Cork, Ireland5 School of Physics and Astronomy, University of St. Andrews, North Haugh, St. Andrews, Fife KY16 9SS, UK6 Harvard Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA7 Department of Physics and Astronomy, The Open University, Walton Hall, Milton Keynes, MK7 6AA, UK8 University of Toledo, Dept. of Physics and Astronomy, Toledo, OH 43606, USA9 Pulkovo Observatory, 196140 Saint-Petersburg, Russia

10 Observatoire de Strasbourg,11 rue de l’Universite, 67000 Strasbourg, France11 Physics Department, University of Crete, P.O. Box 2208, GR-71003, Heraklion, Greece12 Foundation for Research and Technology–Hellas, GR-71110, Heraklion, Greece13 Dept. of Astronomy, San Diego State University, San Diego, CA 92182, USA

Received 7 March 2002; accepted 1 July 2002

Abstract. We present a compilation of spectroscopic observations of the sgB[e] star CI Cam, the optical counter-part of XTE J0421+560. This includes data from before, during, and after its 1998 outburst, with quantitativeresults spanning 37 years. The object shows a rich emission line spectrum originating from circumstellar material,rendering it difficult to determine the nature of either star involved or the cause of the outburst. We collate allavailable pre-outburst data to determine the state of the system before this occurred and provide a baseline forcomparison with outburst and post-outburst data. During the outburst all lines become stronger, and hydrogenand helium lines become significantly broader and asymmetric. After the outburst, spectral changes persist forat least three years, with Fe ii and [N ii] lines still a factor of ∼ 2 above the pre-outburst level and He i, He ii,and N ii lines suppressed by a factor of 2–10. We find that the spectral properties of CI Cam are similar to othersgB[e] stars and therefore suggest that the geometry of the circumstellar material is similar to that proposed forthe other objects: a two component outflow, with a fast, hot, rarefied polar wind indistinguishable from that of anormal supergiant and a dense, cooler equatorial outflow with a much lower velocity. Based on a comparison ofthe properties of CI Cam with the other sgB[e] stars we suggest that CI Cam is among the hotter members of theclass and is viewed nearly pole-on. The nature of the compact object and the mechanism for the outburst remainuncertain, although it is likely that the compact object is a black hole or neutron star, and that the outburst wasprecipitated by its passage through the equatorial material. We suggest that this prompted a burst of supercriticalaccretion resulting in ejection of much of the material, which was later seen as an expanding radio remnant. Theenhanced outburst emission most likely originated either directly from this supercritical accretion, or from theinteraction of the expanding remnant with the equatorial material, or from a combination of both mechanisms.

Key words. stars:individual(CI Cam) – stars:emission line, B[e] – stars:radio emission stars:CI Cam – binaries:close

Send offprint requests to: R. I. Hynes, e-mail:[email protected]

? Guest Observer, McDonald Observatory, University ofTexas at Austin.

1. Introduction

On 1998 April 2 Smith et al. (1998) reported anRXTE All-Sky Monitor (ASM) detection of a bright,rapidly rising X-ray transient designated XTE J0421+560.Subsequently, Marshall et al. (1998) used the RXTE

2 R. I. Hynes et al.: Spectroscopic observations of CI Cam

Proportional Counting Array (PCA) to refine the best fitposition with an error circle of 1 arcmin radius. The bright(V ∼ 11) B[e] star CI Cam(=MWC 84) was found to lienear to the centre of this error circle. Spectroscopic obser-vations by Wagner et al. (1998) on 1998 April 3 revealeda rich emission line spectrum, similar to that reported byDownes (1984; see Sect. 2), but with the presence of He iiemission features. These features had not been reportedin previous spectra, and so by analogy to other X-ray bi-naries Wagner et al. (1998) proposed it to be the opticalcounterpart of XTEJ0421+560. Photometric observationsof the source at this time (e.g. Robinson et al. 1998, Garciaet al. 1998, Hynes et al. 1998) showed that CI Cam wassome 2–3mag brighter than had previously been reported.

Hjellming & Mioduszewski (1998a) reported the detec-tion of a transient 19mJy radio source at 1.4GHz, cor-responding to the optical position of CI Cam on 1998April 1, thus confirming the identification of CI Camas the optical counterpart. Subsequent observation ofrapid radio variability established that the radio emis-sion was of non-thermal (synchrotron) origin (Hjellming& Mioduszewski 1998b). Long term observations indicatethat after the initial flare the radio emission underwentan unusually slow decay, with a 15GHz flux of ∼ 1.5mJyabout 40 months after the initial outburst (Pooley, priv.comm.). High spatial resolution maps obtained after theoutburst indicated the presence of a clumpy ejection neb-ula (Mioduszewski et al., in preparation). These ejectaexpand at ∼ 1.0 − 1.5mas d−1, corresponding to an ex-pansion velocity ∼ 5000km s−1 for a distance of 5 kpc.

Given the distance estimates (and optical luminosityimplied) for CI Cam (e.g. log L/L� ≥ 4.86; Clark et al.2000; Robinson, Ivans & Welsh 2002) it is clear that ifit is a binary, as is likely, it is a high mass X-ray bi-nary (HMXB). However CI Cam does not sit comfortablywithin the traditional divisions of HMXB mass donorsinto classical Be stars (∼ 70 per cent) and OB supergiants(∼ 30 per cent). While its luminosity suggests that it be-longs to the later subset, such systems are typically shortperiod binaries which accrete via Roche lobe overflow ordirect wind fed accretion producing persistent X-ray emis-sion (typically modulated at the orbital period). The pres-ence of a rich emission line spectrum including forbiddenlines, and a near IR excess due to hot dust (the observa-tional criteria for the B[e] phenomenon; Allen & Swings1976; Lamers et al. 1998) also mark a distinction from theother supergiant HMXB systems. Among the stars show-ing the B[e] phenomenon, the high luminosity of CI Cammakes it a Galactic counterpart to the Magellanic Cloudsupergiant B[e] stars (sgB[e] stars) and we will refer to it assuch for the rest of this work. CI Cam therefore appearsto be the first bona fide sgB[e] star HMXB known, al-though direct evidence for binarity has proven elusive anda chance encounter of a compact object with the sgB[e]star although highly unlikely, cannot be ruled out.

This work presents a compilation of spectroscopy ob-tained before, during and after the 1998 outburst. A com-panion photometric compilation has been presented by

Clark et al. (2000). Some of the outburst data includedhere has previously been presented by Barsukova et al.(1998) and Barsukova et al. (2002). In Sect. 2 we sum-marise the available pre-outburst data, including archivaldata with quantitative spectroscopy spanning ∼ 30 yearsbefore the X-ray outburst and additional unpublished pre-outburst spectra. From these we identify the typical pre-outburst strengths of spectral lines and discuss their sta-bility. We then in Sect. 3 describe a series of new spec-tra running from a few days after the X-ray outburst to∼ 3 years later. In Sect. 4 we discuss extinction and dis-tance estimates for the system and in Sect. 5 we reviewwhat is known about the mass donor star. Sect. 6 examinesthe changes in the continuum flux distribution and Sect. 7the spectral lines and how these evolve through the out-burst. Sect. 8 tests for the presence of shorter timescalevariability. Finally in Sect. 9 we will discuss how all ofthese clues can help us build a picture of the nature of thesystem and the outburst mechanism and in Sect. 10 wesummarise our conclusions.

2. Pre-outburst observations

Since CI Cam is bright and has a strong emission linespectrum, there exists a relatively large data set datingfrom 1933 to the present day. We present a summary ofthese observations, which include previously unpublishedoptical spectra from 1987 and 1994, to quantify an average‘quiescent’ state for CI Cam. Table 1 summarises the pre-outburst equivalent widths.

2.1. Published archival spectroscopy

The first observations of CI Cam were made by Merrill &Burwell (1933) and showed strong H i Balmer and weakFe ii emission. He i was seen to be strongly in emissionbut He ii was absent. The first quantitative spectroscopicobservations of CI Cam date from Chkhikvadze (1970).Spectra obtained during the period 1964–67 show strongH i and He i emission lines with no evidence of He ii emis-sion. Fe ii lines were also in emission, the strongest beingmultiplets 27, 28, 37 and 38; there was no evidence for theMg ii 4481 A line. Based on the strength of the Balmeremission lines Chkhikvadze (1970) suggested a spectralclass of O6–B0 for CI Cam.

Allen & Swings (1976) describe further spectroscopicobservations of CI Cam (although no date for the obser-vations is provided). They note numerous emission linesof He i, Fe ii and Si ii. [N ii] lines at 5755 and 6584 A werepresent (the latter in the wings of the very strong Hα line);[O i] lines at 6300 & 6363 A are possibly also detected. Anestimate for the density of the circumstellar envelope ofNe > 106 − 107 cm−3 was derived from the strength ofthe [N ii] lines. Higher excitation lines such as [O iii] and[N iii] appeared to be absent.

A 4000–7000A spectrum was obtained by Downes(1984) in 1984 January and once again was dominatedby strong H i, He i and Fe ii lines.

R. I. Hynes et al.: Spectroscopic observations of CI Cam 3

Miroshnichenko (1995) describes observations madebetween 1986 September and 1987 December. The spec-trum was dominated by strong H i and He i lines, with nu-merous weak Fe ii lines also present (as was C ii emissionat 4267 & 7234A). The H i Balmer lines were symmetricaland single peaked, with wings extending to ±250km s−1

for Hα and Hβ.

Finally, Jaschek & Andrillat (2000) report observa-tions from 1992 and 1998, the latter only two monthsbefore the outburst. 450 emission lines were observed,∼ 55percent from Fe ii, the remainder including H i, He i,O i, N i, Si ii, Mg ii, [O i], [N ii], and [Fe ii]. The Balmerlines show a very steep Balmer decrement. O i 8446 A isunusually strong, likely due to Lyβ fluorescence. Very fewdifferences between the 1992 and 1998 epochs were seen.Since the latter spectra represent the highest quality pre-outburst spectra, as well as the closest to the X-ray out-burst, we reanalysed them to establish a quantitative pre-outburst baseline.

2.2. Unpublished spectra taken at the SpecialAstrophysical Observatory

Three further unpublished observations of CI Cam wereobtained by A. Miroshnichenko at the 6m BTA telescopeof the Special Astrophysical Observatory of the RussianAcademy of Sciences (SAO RAS) on 1987 April 5 (∼4460–5000A) and 1994 January 21 (∼4000–4950 & ∼5700–6800A).

The observations were made with a medium-resolutionspectrograph SP-124 and a 1024-element one-dimensionalphoto-electric detector (Drabek et al. 1986). Data reduc-tion was performed with the sipran software developedat SAO RAS; no attempt at flux calibration was made.Due to the poor dynamic range of the scanner it was pos-sible to obtain either profiles from the strong lines on anunderexposed continuum, or a properly exposed contin-uum with saturated strong lines. Of the data presentedhere, the 1987 spectrum was exposed to search for weaklines, while the two spectra obtained in 1994 were ob-tained to determine the line profiles of the stronger emis-sion lines. Consequently the continuum (and weaker lines)were poorly exposed for these spectra. We have chosen toinclude an equivalent width for the lines identified fromthe 5700-6700A spectrum, although we caution that thecontinuum level was uncertain in this spectrum.

2.3. Archival spectra from Haute-Provence

We obtained the 1998 spectra reported by Jaschek &Andrillat (2000) from the archive of the Observatoire deHaute-Provence (OHP). These spectra were obtained withthe Aurelie spectrograph on the 1.52m telescope using aThomson array of 2026 photodiodes and a 300 line mm−1

grating, yielding a resolution of about 1.3 A.

Line 1964–71 1986–72 19873 19943 19984

Hδ 4100 4.9 – – 5.1 –Hγ 4340 13.4 10.5 – 10.9 10.5Hβ 4861 53 65 57 44 46.5Hα 6562 – 241 – (397) > 300

He i 4026 4.0 – – 3.6 –He i 4471 10.6 – – 10.7 9.6He i 4713 9.3 – 8.0 8.8 8.0He i 5875 93 – – (106.7) 73.6He i 6678 – – – (54.5) 53.2He i 7065 – – – – 89.4He i 7281 – – – – 34.0

[N ii] 5755 – – – – 3.5

Table 1. Pre-outburst equivalent widths, in A, of selectedstrong emission lines. Minus signs have been dropped for con-venience as all lines are in emission. He i 4921 and 5015 A havebeen omitted as these are unresolved from strong Fe ii lines.Sources are 1Chkhikvadze (1970), 2Miroshnichenko (1995),3Section 2.2, 4Jaschek & Andrillat (2000) and Section 2.3.Lines with no value given are not reported, outside the spec-tral range or too weak to reliably measure. Bracketed mea-surements were obtained from spectra with a poorly exposedcontinuum rendering the equivalent widths uncertain.

2.4. The ‘average’ quiescent spectrum

We plot the pre-outburst optical spectrum (from 1998OHP data) in Fig. 1a and in Table 1 we compile the equiv-alent widths of the strongest emission lines in quiescence.Coverage is clearly patchy, but no line exhibits dramaticvariations in equivalent width. In particular the relativestrengths of Balmer and He lines remain roughly the same.Downes (1984) does not include equivalent widths, but therelative line strengths plotted in his spectra (e.g. Hγ toHe i 4471) look very similar to the results in Table 1, andalso to Fig. 1a. It therefore appears that the pre-outburstspectrum was stable for at least ∼ 30 years before theoutburst, although large gaps in coverage are obviouslypresent. As discussed in Sect. 7, the post-outburst spec-trum is clearly different to this, but it remains to be seenwhether the differences represent a delayed return to thepre-outburst quiescent state, or a new different quiescentstate.

3. Outburst and post-outburst observations

We next present a series of spectroscopic observations ob-tained during 1998 April and May, covering the outburstand immediate aftermath. Spectroscopy taken during thistime indicates the presence of high excitation lines of He iiat the time of the outburst which fade on a timescale∼ 10 days; similar variability is seen in most other linesobserved during this period. Such behaviour occurs overlonger timescales than the X-ray outburst; XTE observa-tions reveal that the X-ray flux had already dropped by afactor of a hundred by April 4 (Belloni et al. 1999).


Fig. 1. Comparison between normalised spectra from a) before the outburst (1998 January 27–28), b) early in the outburst(1998 April 4), and c) after the outburst (2000 February 6 for λ > 7100 A and 2000 December 1 elsewhere). Major features arelabelled. The strongest groups of Fe ii lines are also indicated, but other species are present in these complexes and Fe ii linesare also found elsewhere.

We obtained both blue and red spectra during the out-burst, thus enabling us to follow the evolution of the out-burst and determine changes in the circumstellar environ-ment during this period. Spectra were obtained from fourtelescopes in the month following the X-ray outburst: the1.5m at the FLWO, the 6m telescope of the SAO RAS, the4.2m WHT on La Palma and the McDonald Observatory2.7m. Subsequent observations obtained during the pe-riod between 1998 May and 2001 October constrain thelong term post-outburst behaviour of the system. All ofthese observations are summarised in Table 2 and moredetails are given in the following subsections.

3.1. F. L. Whipple Observatory observations

A series of spectra were obtained using the 1.5m tele-scope at the F. L. Whipple Observatory (FLWO) and theFAST spectrometer. These observations began soon af-ter the outburst was discovered and continued until theend of 1998. The spectra were obtained with two differentgratings, of 300 and 1200 lines mm−1, giving 3.0 and 1.1Aresolution respectively.

These observations included a series of 370 2 s spectrawith ∼ 1.1A resolution and a dispersion of 0.4 A pix−1,covering 6040–7035A which were obtained on the night of1998 April 2. We searched for shifts in the velocities of theemission lines by cross-correlating all the spectra against


Date Telescope UT Start UT End Number Wavelength ResolutionRange(A) (A)

Pre-outburst05/04/87 SAO RAS 6-m 23:12 23:50 1 4460–5005 1.106/04/87 SAO RAS 6-m 00:20 00:45 1 6620–7290 2.121/01/94 SAO RAS 6-m 17:38 18:06 1 4000–4900 2.221/01/94 SAO RAS 6-m 18:15 18:28 1 5700–6700 2.126/01/98 OHP 1.52-m 21:37 22:20 1 4060–4930 1.327/01/98 OHP 1.52-m 21:16 22:46 1 4860–5730 1.327/01/98 OHP 1.52-m 23:14 00:44 1 6250–7110 1.328/01/98 OHP 1.52-m 01:04 02:34 1 4060–4930 1.328/01/98 OHP 1.52-m 23:02 00:32 1 8040–8900 1.329/01/98 OHP 1.52-m 01:08 02:38 1 5560–6430 1.329/01/98 OHP 1.52-m 03:00 04:30 1 7060–7930 1.3

Outburst03/04/98 FLWO 1.5-m 02:41 02:44 3 3580–7450 3.003/04/98 FLWO 1.5-m 02:47 04:19 370 6040–7035 1.104/04/98 SAO RAS 6-m 17:07 17:11 2 3700–6130 4.004/04/98 SAO RAS 6-m 17:14 17:15 1 4990–7460 4.005/04/98 SAO RAS 6-m 15:33 15:35 1 3700–6130 4.005/04/98 SAO RAS 6-m 15:40 15:41 1 4990–7460 4.006/04/98 SAO RAS 6-m 15:47 16:20 2 3700–6130 4.006/04/98 SAO RAS 6-m 15:52 16:04 3 4990–7460 4.009/04/98 WHT 4.2-m 21:15 21:46 2 3870–6110 0.110/04/98 WHT 4.2-m 20:50 21:48 3 3870–6110 0.111/04/98 WHT 4.2-m 20:49 21:51 5 3870–6110 0.118/04/98 McDonald 2.7-m 02:47 05:00 61 6190–6900 1.218/04/98 FLWO 1.5-m 03:08 03:28 11 3630–7490 3.019/04/98 McDonald 2.7-m 02:26 04:07 65 6190–6900 1.219/04/98 FLWO 1.5-m 02:54 03:10 2 3740–4740 1.119/04/98 FLWO 1.5-m 03:08 03:18 3 4310–5310 1.119/04/98 FLWO 1.5-m 03:27 03:32 4 6040–7030 1.120/04/98 McDonald 2.7-m 02:12 04:54 80 6190–6900 1.219/04/98 SAO RAS 6-m 14:58 15:15 25 3700–6130 8.019/04/98 SAO RAS 6-m 15:19 15:27 2 4990–7460 8.0

Table 2. Summary of spectroscopic observations, both pre-, during, and post-outburst which were quantitatively analysed forthis work. In some cases multiple spectra from the same night covering the same wavelength range have been merged on oneline; in this case start and end times refer to the first and last spectrum and do not indicate total exposure time. Continued onthe next page.

a single template. The largest shifts are ±12km s−1, therms in the velocities is 3.6 km s−1. Given the high airmass(from 1.5 to 2.0) and the 2.0 arcsec slit, these variationsare likely due to variable illumination of the slit ratherthan intrinsic motions in the emission lines, as the velocityresolution was about 50 km s−1. Nonetheless, we searchedfor periodicities in the velocities using the algorithm ofStellingwerf (1978), but found nothing significant. We willfurther discuss evidence for variations in line strengths andprofiles in Sect. 8.

3.2. Special Astrophysical Observatory observations

Outburst observations spanning 1998 April to Maywere made using the 6 m BTA telescope of theSAO RAS. A medium resolution spectrograph, SP–124 (at the Nasmyth–1 focus), was used with a

Photometrix 1024×1024 CCD detector and the B1 grat-ing (600 lines mm−1 ) providing a spectral dispersion of2.4 A pixel−1 in two overlapping spectral ranges 3700–6100 and 5000–7500A. The slit width employed was either1 arcsec or 2 arcsec, depending on seeing conditions, pro-viding a spectral resolution of ∼ 4 A or ∼ 8 A respectively.Primary reduction of the CCD spectra including correc-tions for the dark current, debiasing and sky subtractionwas accomplished with midas (1996 Nov. version). He–Ne–Ar lamps were used for wavelength calibration.

3.3. WHT/UES observations

CI Cam was observed at high resolution using the UtrechtEchelle Spectrograph (UES) on the William HerschelTelescope (WHT) on La Palma on 1998 April 9–11.Spectra were extracted using a simple (non-optimal) ex-



Post-outburst16/05/98 SAO RAS 6-m 18:08 18:21 3 3700–6130 4.016/05/98 SAO RAS 6-m 18:29 18:40 2 4990–7460 4.020/07/98 WHT 4.2-m 05:43 06:01 4 3500–7000 3.121/07/98 WHT 4.2-m 05:57 06:06 2 4800–5200 0.415/09/98 FLWO 1.5-m 09:49 09:53 3 3660–7530 3.017/09/98 FLWO 1.5-m 10:46 10:50 2 3660–7530 3.018/09/98 FLWO 1.5-m 11:43 11:47 2 3660–7530 3.019/09/98 FLWO 1.5-m 11:42 11:44 3 3660–7530 3.021/09/98 FLWO 1.5-m 08:54 08:56 3 3660–7530 3.023/09/98 FLWO 1.5-m 11:42 11:46 2 3660–7530 3.024/09/98 FLWO 1.5-m 11:41 11:45 2 3660–7530 3.029/09/98 FLWO 1.5-m 11:57 11:58 3 3630–7490 3.030/09/98 FLWO 1.5-m 08:37 08:39 3 3660–7530 3.014/10/98 FLWO 1.5-m 10:19 10:22 2 3650–7520 3.015/10/98 FLWO 1.5-m 10:55 10:58 2 3650–7520 3.016/10/98 FLWO 1.5-m 10:44 10:47 2 3650–7520 3.023/10/98 FLWO 1.5-m 10:49 10:51 3 3580–7450 3.029/10/98 FLWO 1.5-m 09:03 09:07 2 3650–7520 3.001/11/98 Asiago 1.82-m 02:20 02:35 1 6360–8060 3.312/11/98 FLWO 1.5-m 10:02 10:13 3 3630–7500 3.013/11/98 FLWO 1.5-m 11:41 11:44 3 3630–7500 3.014/11/98 FLWO 1.5-m 10:02 10:06 3 3630–7500 3.015/11/98 FLWO 1.5-m 08:14 08:17 3 3630–7500 3.016/11/98 FLWO 1.5-m 10:41 10:44 3 3630–7500 3.017/11/98 FLWO 1.5-m 11:08 11:10 4 3630–7500 3.018/11/98 FLWO 1.5-m 09:06 09:18 4 3630–7500 3.019/11/98 FLWO 1.5-m 09:40 09:44 3 3660–7530 3.021/11/98 FLWO 1.5-m 09:31 09:34 3 3660–7530 3.022/11/98 FLWO 1.5-m 08:35 08:39 3 3660–7530 3.023/11/98 FLWO 1.5-m 09:04 09:07 3 3660–7530 3.024/11/98 FLWO 1.5-m 10:24 10:27 3 3660–7530 3.025/11/98 FLWO 1.5-m 07:56 08:01 3 3660–7530 3.026/11/98 FLWO 1.5-m 07:47 07:51 3 3660–7530 3.027/11/98 FLWO 1.5-m 08:23 08:27 3 3660–7530 3.030/11/98 FLWO 1.5-m 11:17 11:22 3 3660–7530 3.0

Table 2. (continued)

traction with sky subtraction using echomop software.Cosmic rays were removed by hand.

3.4. McDonald Observatory observations

Intermediate dispersion spectroscopy was obtained on thenights of 1998 April 18–20 using the Large CassegrainSpectrometer (LCS) on the McDonald Observatory 2.7mtelescope. Each night involved a series of observationsspanning about 2 hours to search for variability; this willbe discussed further in Sect. 8. Reduction and extractionwere performed using standard iraf tasks. Observationsof neon arc lines were performed at 20min intervals toprovide wavelength calibration. Corrections were made tothe interpolated wavelength solution using night sky emis-sion lines. As for the FLWO observations, no significantwavelength variations were found.

3.5. WHT/ISIS observations

Intermediate and high dispersion spectroscopy was ob-tained on several nights from 1998–2001 using the ISISdual-beam spectrograph on the WHT. Most exposuresused the blue arm with the EEV10 or EEV12 CCD.The second spectrum from 2001 April 28 used the redarm with the TEK4 CCD. Slit widths were 0.7–1.0arcsec.Reduction and extraction were performed using standardiraf tasks. Copper-neon and/or copper-argon lamps wereobserved to provide wavelength calibration. No flux cali-bration of these spectra was attempted.

3.6. Loiano & Asiago Observatory observations

Post-outburst spectra of the source were obtained in 1999January 5 and 2000 February 6 using the 1.52m G. D.Cassini telescope at the Loiano Observatory (Italy). The



10/12/98 FLWO 1.5-m 05:59 06:03 3 3660–7530 3.012/12/98 FLWO 1.5-m 07:29 07:32 3 3660–7530 3.013/12/98 FLWO 1.5-m 05:29 05:33 3 3660–7530 3.014/12/98 FLWO 1.5-m 09:25 09:29 3 3660–7530 3.019/12/98 FLWO 1.5-m 06:07 06:11 3 3660–7530 3.020/12/98 FLWO 1.5-m 06:22 06:27 3 3660–7530 3.021/12/98 FLWO 1.5-m 07:35 07:39 3 3660–7530 3.022/12/98 FLWO 1.5-m 06:08 06:12 3 3660–7530 3.023/12/98 FLWO 1.5-m 06:11 06:15 3 3660–7530 3.025/12/98 FLWO 1.5-m 08:32 08:36 3 3660–7530 3.026/12/98 FLWO 1.5-m 04:52 04:56 3 3660–7530 3.027/12/98 FLWO 1.5-m 05:14 05:17 3 3660–7530 3.028/12/98 FLWO 1.5-m 03:14 03:18 3 3660–7530 3.003/01/99 OHP 1.52-m 00:26 01:55 1 8040–8900 1.303/01/99 OHP 1.52-m 03:30 05:30 1 4060–4930 1.303/01/99 OHP 1.52-m 20:05 21:35 1 4860–5730 1.303/01/99 OHP 1.52-m 22:11 23:41 1 5560–6430 1.304/01/99 OHP 1.52-m 00:04 01:34 1 6250–7110 1.304/01/99 OHP 1.52-m 02:33 04:33 1 7060–7930 1.305/01/99 Loiano 1.52-m 21:17 21:27 1 6360–8220 3.329/01/00 Calar Alto 3.5-m 23:44 23:49 1 3700–6800 6.106/02/00 Loiano 1.52-m 21:11 21:41 1 3500–5400 5.506/02/00 Loiano 1.52-m 20:33 21:06 2 3530–8830 8.319/07/00 Skinakas 1.3-m 02:30 02:35 1 5550–7550 4.516/10/00 Skinakas 1.3-m 01:37 01:40 1 5250–7300 4.501/12/00 WHT 4.2-m 20:22 20:37 3 3700–7200 3.601/12/00 WHT 4.2-m 21:06 21:37 3 3870–4310 0.328/04/01 WHT 4.2-m 20:53 21:03 1 3600–4030 0.328/04/01 WHT 4.2-m 20:45 21:03 15 6300–6700 0.608/08/01 Skinakas 1.3-m 01:42 01:47 1 5460–7430 4.513/09/01 Skinakas 1.3-m 00:45 00:48 1 5230–7200 4.523/10/01 Bok 2.3-m 10:30 10:46 3 3885–5030 1.8

Table 2. (continued)

telescope was equipped with the Bologna Faint ObjectSpectrograph and Camera (BFOSC) and a 2048×2048Loral CCD. In January 1999 grism #8 was used, giv-ing a resolution of ∼ 3A, and in February 2000 the low-dispersion grisms #3 (blue) and #4 (red) were used. Onefurther spectrum was taken on 1998 November 1st us-ing the 1.82m telescope at Asiago Observatory (Italy),equipped with the Asiago Faint Object Spectrograph andCamera (AFOSC) and a 1024×1024 thinned SITe CCD.This instrument is identical to BFOSC and grism #8 wasused, giving the same resolution.

3.7. Haute Provence observations

Post-outburst spectra obtained from OHP in 1999 usedthe same configuration as the 1998 pre-outburst obser-vations, hence the details are the same as described inSect. 2.3.

3.8. Calar Alto observations

One low-resolution flux-calibrated spectrum was taken on2000 January 29 using the blue arm of the TWIN spec-trograph on the 3.5m telescope at Calar Alto (Spain) anda 2000×800 thinned SITe CCD.

3.9. Skinakas Observatory observations

Spectra taken on 2001 July 19 & October 16 weretaken with the 1.3m f/7.7 Ritchey-Chretien telescope atSkinakas Observatory (Crete, Greece). This was equippedwith a 2000×800 ISA SITe CCD and a 1302 linesmm−1

grating (using a 80µm slit), giving a dispersion of∼1 A pixel−1.

3.10. Steward Observatory observations

Finally three spectra were obtained using the Bok 2.3mtelescope of the Steward Observatory with the Boller andChivens long slit spectrograph and a 1200×800 CCD. Afirst order 1200 linesmm−1 grating was used and there


were no filters. Flux calibration was not reliable as therewere significant slit losses.

4. Interstellar extinction and distance estimates

There remain large uncertainties about the interstellar ex-tinction to CI Cam, and even more so concerning the dis-tance. A number of extinction estimates have been madebut these are complicated by the uncertainty about howmuch of the measured extinction is intrinsic to the source;different methods may be more or less sensitive to this de-pending on what is actually measured. Consequently it isof value to compare as many independent methods as pos-sible. We discuss only recent estimates as these are likelyto use higher quality data and/or more sophisticated mod-els, superseding earlier estimates (e.g. Chkhikvadze 1970).

A number of attempts have been made to fit the broadband optical–far IR spectral energy distribution (SED)using a model including extinction as a free parameter.Belloni et al. (1999) fitted the SED using a Kurucz at-mosphere and a dust model and derived E(B − V ) =1.18 ± 0.04mag and hence AV ∼ 4.4mag. Clark et al.(2000) used a similar model to derive AV ∼ 3.71mag.Zorec (1998) used a model in which the interstellar ex-tinction and distance were constrained to be consistentrather than independent, and estimated d ∼ 1.75kpc,E(B − V )ism ∼ 0.8mag and AV,local ∼ 2.4mag, imply-ing a total extinction somewhat higher than the other twoestimates.

Orlandini et al. (2000) used emission line ratios to esti-mate E(B−V ) = 1.54mag (from He lines) or E(B−V ) =1.02mag (from H lines). This is based on theoretical pre-dictions of the ratios of the strongest lines in CI Cam. InBe stars and other early OB type stars with dense winds,and likely also in sgB[e] stars, these lines are subject toeffects of non-local thermodynamic equilibrium (NLTE)so this method will not be very reliable.

Robinson et al. (2002) use the 2175 A interstellar ab-sorption feature in post-outburst HST data to measureE(B− V ) = 0.85± 0.05mag and AV = 2.3± 0.3mag; thelarge error on AV reflects the uncertainty in the choice ofextinction curve parameterised by RV = AV /E(B − V ).

We can attempt to estimate the Na D equivalent width,although as noted by Munari & Zwitter (1997) this isinsensitive for reddened objects as it tends to saturate.It is also complicated by the presence of Na D emission.Fortunately, our echelle spectra obtained in outburst re-solve the Na D components as sharp absorptions withinthe emission line (Fig. 2a). Reconstructing the unabsorbedline profile is obviously uncertain, but assuming it is sym-metric we can estimate that the equivalent widths of thetwo Na D2 (5890 A1) components observed are 0.36 and0.52 A. The latter component does appear saturated, sothis only provides a lower-limit to the reddening. If the

1 We note that Munari & Zwitter 1997 incorrectly label the5890 A line as NaD1, and 5896 A as NaD2. Their work is,however, internally consistent; Munari, priv. comm.

Fig. 2. a) NaD line profiles observed during outburst (1998April 9). The two interstellar absorption components areclearly visible; the longer wavelength one is saturated. Thedashed line shows the reconstructed NaD2 line profile assum-ing it is symmetrical. b) Closeup of the absorption componentsindicating the velocities with respect to the LSR. The struc-ture in the two lines is consistent. In this direction the LSRvelocity is expected to become more negative with distance, sothe distance increases to the left. A quantitative distance scaledepends on the assumed rotation curve of the Galaxy.

profile is actually asymmetric with the same extendedblue wing seen in many other lines then this will in-crease the inferred equivalent width, implying a redden-ing further above our lower limit. Using the calibrationof Munari & Zwitter (1997) for each of the two compo-nents we derive a lower limit for the combined reddeningof E(B − V ) >∼ 0.5mag.

Our spectra also reveal several diffuse interstellarbands (DIBs). Our principal reddening calibrator is theλ5780 band, using the calibration of Herbig (1993). Usingthe WHT/UES spectra we measure an EW of 320±30mAimplying E(B − V ) = 0.64 ± 0.06mag. The DIB at5797 A is too blended with emission lines to reliably mea-sure but we can approximately estimate 140± 30mA im-plying E(B − V ) = 0.9 ± 0.2mag (Herbig 1993). For


5705 A we measure 110 ± 20mA implying E(B − V ) =0.49 ± 0.09mag (Herbig 1975). We also use an averageof several WHT/ISIS and Skinakas spectra to estimatean equivalent width of 180 ± 30mA for λ6203 implyingE(B − V ) = 0.62 ± 0.11mag (Herbig 1975), and a cen-tral depth of 0.90–0.97 for the very broad λ4428 featureimplying 0.3 < E(B − V ) < 1.1 (Krelowski et al. 1987).The 6284 A line may also be present but is complicated byTelluric absorption. All of the reliable DIB measurementsare consistent with E(B− V ) ∼ 0.65mag as suggested byour principal calibrator, 5780 A.

To summarise, methods based on interstellar absorp-tion features (2175 A, NaD, DIBs) all suggest E(B−V ) ∼0.5 − 1.0mag, and hence visual extinction AV ∼ 1.5 −3mag. Methods based on fitting the SED or line ratios,however, favour larger total extinction AV ∼ 4mag. Thiscould arise if the absorption features are only producedin the interstellar medium, and not by local dust extinc-tion, as a consequence of different chemical compositionsand/or grain sizes. The SED methods are all model de-pendent, however, and it may be that the models used areeither incorrect or incomplete. This leaves us with someambiguity about how to deredden spectra of CI Cam. Thehigher values inferred from SED modelling are most rele-vant, in the sense that they directly measure the distortionof the broad band spectrum, but they are also model de-pendent. Measurements based on absorption features aremore objective, but do not measure the same thing.

Distance estimates for CI Cam are plagued by evenmore uncertainty than extinction estimates. A numberof works have derived distances of 1–2 kpc (Chkhikvadze1970; Zorec 1998; Belloni et al. 1999; Clark et al. 2000).All of these, however, involve uncertain assumptions aboutthe luminosity class of the star, or the relation between in-terstellar absorption and distance. Robinson et al. (2002)challenged this conclusion, arguing for a larger distance.Based on spectroscopic similarities to the largest and mostluminous sgB[e] stars they concluded that the distancehad to be larger than 2 kpc. Since the line of sight passeswell above the warped Galactic plane for 2–6 kpc, theyargued that a young object like CI Cam must lie beyondrather than within this distance range. In support of thisthey note that its radial velocity is consistent with a dis-tance of ∼ 7 kpc, assuming differential Galactic rotation.These arguments, however, make the assumptions thatCI Cam is comparable to the most luminous sgB[e] stars,and that it does not have significant peculiar velocity.

More objective constraints on the distance are harderto obtain. We can, however, make one estimate whichis almost completely independent of what CI Cam is.As already described, the Na D lines show sharp ab-sorption components. The velocities of these in the lo-cal standard of rest (LSR) are −4.5 ± 0.6 km s−1 and−34.8±0.4km s−1 for the stronger and weaker componentsrespectively (see Fig. 2b). The stronger component is sat-urated so it could itself involve multiple components withLSR velocities up to −10km s−1. The weaker componentmay also have an additional component in its blue wing at

around −45 km s−1; both the NaD1 and NaD2 lines sug-gest such a feature at this velocity. All of these componentsare somewhat redshifted with respect to CI Cam, whichhas an LSR velocity of −51km s−1 (Robinson et al. 2002).It is therefore unlikely they are associated with circumstel-lar material; much more likely is that they are interstel-lar gas. If so then we expect them to be moving in theplane in near circular orbits following Galactic rotation,and their velocities can be used to estimate their distances,and hence a lower limit on the distance of CI Cam. Thereare obviously uncertainties introduced by non-circular mo-tions and an imperfectly known rotation curve. We esti-mate distances using the range of rotation curves illus-trated by Olling & Merrifield (1998), with the spread invalues giving an estimate of the uncertainty in the mea-surement. The saturated component is clearly associatedwith the Local Arm, with a maximum velocity −10km s−1

corresponding to a maximum distance ∼ 1 kpc. We mightexpect the next feature to correspond to the Perseus Armat a distance∼ 2.5 kpc, but the Perseus Arm is not well de-fined in the direction of CI Cam as our line of sight passeswell above the Galactic plane at that point (c.f. discus-sion by Robinson et al. 2002), so it is unsurprising that itproduces no absorption feature. The next feature out, at−35 kms−1, is at an implied distance of 3.9–5.6kpc, andprobably corresponds to the next spiral arm out, whereour line of sight passes back into the warped outer disc;Cam OB3 (Humphreys et al. 1978) likely belongs to thisarm. If the third weak feature, at −45 kms−1, is real andalso indicative of Galactic rotation, then it suggests aneven greater minimum distance of 6–8 kpc, dependent onthe assumed rotation curve.

An alternative empirical approach is to use the radialvelocity maps of Brand & Blitz (1993), derived from H iiregions and reflection nebulae. This will take account ofnon-circular motions. If anything, these suggest an evenlarger distance. The coverage in the direction of CI Cam issparse, but representative distant objects from their sam-ple (S208, S211 and S212) have LSR velocities between-30 and -38km s−1 and distance estimates of 5.9–7.6kpc.

The distance implied for CI Cam by the interstellarfeatures is thus large; it is at least 4 kpc, and may wellbe beyond 6 kpc. This conclusion is essentially consistentwith that derived independently by Robinson et al. (2002)using different methods.

If the closest distance is adopted it is possible thatCI Cam could be associated with Cam OB3, possibly as arunaway object. It lies ∼ 2.◦7 from the centre of the associ-ation; the members identified by Humphreys (1978) lie atup to 1.◦2 from the centre. The distance modulus adoptedby Humphreys (1978) was 12.6 (3.3 kpc). CI Cam there-fore lies somewhat outside the association, and appearsto be further away, but we cannot rule out the possibilitythat it is a runaway from Cam OB3. Alternatively it couldlie beyond the spiral arm containing Cam OB3.

Where it is necessary to assume a value for the redden-ing in what follows, we either take E(B − V ) = 1.3mag(AV = 4mag) or a range of 0.65 < E(B − V ) < 1.4


(2.0 < AV < 4.4). For the distance, we follow Robinson etal. (2002) in adopting 5 kpc as a representative estimate,although we agree that it could be somewhat larger thanthis. Where a lower limit is more appropriate we assumed > 4 kpc.

5. The nature of the components of CI Cam

CI Cam clearly shows the observational characteristics ofthe B[e] phenomenon (Allen & Swings 1976; Lamers etal. 1998): strong Balmer emission lines (e.g. Fig. 1), lowexcitation permitted lines (e.g. Fig. 1), optical forbiddenlines of [Fe ii] and [O i] (e.g. Fig. 12a) and a strong in-frared excess (Clark et al. 2000); indeed CI Cam has longbeen considered a B[e] star and was included in the sam-ple of Allen & Swings 1976. In view of the distance es-timates discussed above CI Cam clearly falls within thesgB[e] sub-class. The primary characteristic of an sgB[e]star, a luminosity of above 104 L� (Lamers et al. 1998) issatisfied for any distance above 0.8 kpc and for the lowerlimit of 4 kpc that we have argued for above, this rises to105.4 L�. Robinson et al. (2002) also note spectroscopicsimilarities to the most luminous sgB[e] stars, consistentwith this. The other known types of stars showing theB[e] phenomenon are the pre-main sequence Herbig Ae/Be(HAeB[e]) stars, compact planetary nebula B[e] (cPNB[e])stars and symbiotic B[e] (symB[e]) stars. HAeB[e] starstypically have luminosities of <∼ 104.5 L�, are associatedwith star forming regions and sometimes show evidenceof infall, e.g. inverse P Cygni profiles. CI Cam is more lu-minous than this and exhibits a steep decrease in the IRflux longward of ∼ 10µm (Clark et al. 2000), indicativeof the absence of the cold dust that would be expectedaround an HAeB[e] star. Lamers et al. (1998) did suggestthat CI Cam could be a cPNB[e] star, but as these objectshave lower luminosities (<∼ 104 L�) this is also ruled outby the distance estimates discussed in Section 4. SymB[e]stars are binaries also containing a cool giant which is usu-ally seen in the red or infrared spectrum. Some authorshave identified CI Cam as a symbiotic (e.g. Barsukova etal. 2002), but late type features are not seen in CI Cam(except for the report of Miroshnichenko 1995 which wasnot corroborated by any other observations.) We believethat both the observed spectrum and the high luminositytherefore identify CI Cam as an sgB[e] star, as argued byRobinson et al. 2002.

The sgB[e] class itself includes a range of objects withlikely spectral types spanning B0–B9, and luminositiesfrom 104.2–106.1 L� (Lamers et al. 1998). It is thereforeof interest to attempt a more precise classification forCI Cam. We might hope that the detection of photosphericlines from the sgB[e] star would be of great help. In thehighest resolution post-outburst spectra of CI Cam, broadabsorption wings are seen around higher order Balmerlines, most prominently Hδ and Hε (Fig. 3). These maywell be photospheric absorption lines from the sgB[e] star.Unfortunately we cannot obtain any detailed diagnosticsfrom these lines because of the heavy contamination we see

Fig. 3. Absorption wings surrounding Hδ and Hε lines, from2000 December 1 WHT data. Unmarked lines are mostly Fe ii.

in the wings of the lines (both in terms of other lines andalso probably residual wind emission); hence it is impos-sible to quantify the underlying photospheric spectrum.Indeed, the photospheric spectrum may be poorly definedfor a star with a very high mass loss rate for which a bonafide photospheric radius is hard to determine.

While the spectral type of the star is difficult to de-termine from the optical spectrum due to heavy contam-ination by the circumstellar material, the detection ofP Cygni profiles in the UV resonance lines with absorp-tion to ∼ 1000km s−1 (Robinson et al. 2002) is more use-ful. By analogy with other sgB[e] stars we expect that theUV P Cygni profiles arise from a hot, polar wind whichis similar to that of normal hot supergiants. The outflowvelocity implied for CI Cam is directly comparable to thatseen for other early B supergiants. The Si iv 1394, 1402Adoublet functions as a powerful probe of temperature andluminosity for B stars (Walborn, Parker, & Nichols 1995).Early-B supergiants show a strong P Cygni wind profile,which evolves into pure absorption for mid-B stars andis absent in late B stars of all luminosity classes – theP Cygni profiles therefore suggest an early B classifica-tion. A similar conclusion can be drawn from the presenceof P Cygni profiles for the C iv 1549,51 A doublet whichis seen for supergiants of type B4 or hotter (and is seenin absorption for hot B0–2 dwarfs and giants and coolersupergiants).

By comparison of the optical emission lines withother sgB[e] stars and luminous blue variables (c.f.Miroshnichenko 1996), we can further narrow the spectraltype to B0–B2, as earlier spectral types show strongerHe ii emission than the very weak pre-outburst featurethat we see, and later types show He i in absorption ratherthan emission.

To summarise, the ‘normal’ star is an sgB[e] star,with likely spectral type B0–B2 and a luminosity of atleast 105.4 L�, placing it among the hotter, more luminoussgB[e] stars (c.f. Robinson et al. 2002).


Fig. 4. Flux calibrated spectra obtained from FLWO early inoutburst and after it. These have been dereddened using theFitzpatrick (1999) extinction curve assuming AV = 4.0. Inoutburst, the lines become much stronger and broader, Balmerjump emission appears, and the continuum becomes redder.Note that these are long exposure spectra to ensure that thecontinuum is well defined, so the stronger emission lines aresaturated.

The nature of the compact object implicated in theoutburst is even more uncertain than that of the nor-mal star. It is widely assumed from the X-ray and γ-rayoutburst observed that this must be a black hole or neu-tron star (e.g. Belloni et al. 1999; Robinson et al. 2002).Orlandini et al. (2000), however, have argued for a ther-monuclear runaway on the surface of a white dwarf. Ifthe 4 kpc lower limit on the luminosity is correct, how-ever, then the X-ray outburst was extremely luminous,LX >∼ 2× 1038 erg s−1. This corresponds to the Eddingtonlimit for a neutron star, and for larger distances thena black hole becomes more likely. However, neither theX-ray spectral shape (Orr et al. 1998; Ueda et al. 1998;Revnivtsev et al. 1999; Belloni et al. 1999) nor the lackof rapid X-ray variability (Frontera et al. 1998; Belloni etal. 1999) are typical of accreting black holes or neutronstars. Equally, while there may be a hint of a flaring softcomponent (Frontera et al. 1998; Ueda et al. 1998), theX-ray spectrum is on the whole quite hard, and not domi-nated by a supersoft component as might be expected fora high luminosity accreting white dwarf. Consequently theidentification must remain uncertain.

6. The spectral flux distribution

Clark et al. (2000) considered the broad band spectralenergy distribution from photometric measurements andexamined colour changes during the outburst. They foundthat the source became systematically redder when it wasbrighter. Given the large contribution of emission lines tothe spectrum, however, it is useful to attempt to isolatecontinuum changes, and quantify the line contribution.

In Fig. 4 we show the first outburst spectrum obtained,which fortunately was flux calibrated, together with acomparable post-outburst one. It is clear that the red-der spectrum seen in outburst was not just due to linechanges, but that the underlying continuum is redder. Wewill further discuss the origin of this enhanced continuumemission in Section 9.3. There is also an enhancement athigher frequencies as the Balmer jump appears in emis-sion, which explains why the (U−B) index remained con-stant during outburst (Clark et al. 2000).

In principle, we could also use flux calibrated spec-tra to estimate how much of the flux in a given bandpasscomes from lines. However, in many cases lines are satu-rated in these spectra, so we instead use unsaturated, butuncalibrated spectra. To do this we construct a flux distri-bution by interpolating between the outburst JKT points(Hynes et al. 1998) and the mean post-outburst photom-etry (Clark et al. 2000). We can then multiply such a fluxdistribution by a continuum normalised spectrum and per-form synthetic photometry, iterating until we have a fluxdistribution such that the synthetic photometry matchesthe observed values. This is very crude, but for estimat-ing the contribution of lines within a bandpass, the effectof the assumed flux distribution is only to introduce awavelength dependent weighting within the bandpass, sothe results are weakly sensitive to the flux distribution as-sumed. The contribution from the lines can be estimatedby comparing synthetic photometry performed on the fluxdistribution alone with that performed on the flux distri-bution multiplied by the normalised spectrum. The dif-ferences in magnitudes which we infer in outburst (1998April 4) are ∆B ∼ 0.6, ∆V ∼ 0.6 and ∆R ∼ 0.9, and post-outburst ∆B ∼ 0.08, ∆V ∼ 0.10 and ∆R ∼ 0.35. Whilethe quiescent values are fairly small, and hence the qui-escent photometry is dominated by continuum, the out-burst values can be large; the R band flux, for example, isabout 50 percent line emission, due to the strong Hα line.These values are similar to those quoted by Barsukova etal. (2002).

7. Spectral line behaviour

The spectrum of CI Cam has long been known to berich in emission lines. In outburst, this is even more dra-matic, with most lines showing larger equivalent widths,and Balmer and He i lines completely dominating theirregions of the spectrum (see Fig. 1). A similarly rich out-burst spectrum was seen in the infrared (Clark et al. 1999).We now consider the evolution of selected line strengthsand profiles before, during and after the outburst. We useall the data described earlier together with some measure-ments reported by Wagner et al. (1998) and Orlandiniet al. (2000). Equivalent width lightcurves for a repre-sentative range of lines are collated in Fig. 5; these arediscussed in turn in the following subsections. No statis-tical errors have been indicated as the dominant uncer-tainty is systematic due to contamination from other linesand the uncertain continuum level in such a rich spec-


Fig. 5. Equivalent width evolution of various lines during and after the outburst. All equivalent widths have been normalised topre-outburst levels, shown as a dashed line, except where noted. All points correspond to individual nights except for the 1998Sept–Dec FLWO data for which monthly averages have been plotted as the evolution is slow by this time. The zero-point oftime corresponds to the peak of the X-ray outburst. a) Balmer line evolution. Hα is not plotted because it is so strong as to beunreliable in many spectra. b) He i evolution. Lines at 4921 A and 5015 A are omitted as they are blended with strong Fe ii linesin most spectra. c) The He ii 4686 A evolution. d) Evolution of 5650–5700 A blend. This is likely dominated by N ii 5676,5686at the peak. e) Evolution of Fe ii lines. The spectral lines used were: λλ4233, 4303, 4351, 4385, 4416 (multiplet 27); λλ6432,6516 (multiplet 40); λλ5197, 5234, 5275, 5316, 5325, 5425 (multiplet 49); and λλ6238, 6248, 6416, 6456 (multiplet 74). For eachmultiplet the sum of equivalent widths has been taken. For multiplets 40, 49 and 74 this was normalised to the equivalent sumbefore the outburst. These multiplets are then in approximate agreement by the end of the period observed (around day 1000),so multiplet 27, for which no reliable pre-outburst measurements are possible, was also arbitrarily normalised to agree withthem. The rise between the first few observations is definitely real, being seen in all 12 lines for which measurements could bemade. f) Evolution of [N ii] 5755 A (filled circles).


trum. Consequently the best indicator of the reliability ofa lightcurve is the scatter of the plotted points.

7.1. Hydrogen and helium lines

During outburst H i, He i, and He ii lines all brightendramatically. The increase is much more dramatic thanin the continuum, so equivalent widths rise. The risingphase is not observed, except possibly in He ii 4686 Awhere the earliest observations suggest that the equiva-lent width peaked around 5 days after the X-ray outburst.Comparison with photometry obtained at this time indi-cates that this arises because during this period the He iiflux was approximately constant while the continuum de-cayed. For the H i and He i lines, the fact that these linesare already decaying within two days of the X-ray out-burst indicates that they originate in a region that canrespond to the outburst in less than two days.

There are clearly major differences between the H iand He i behaviours. All the Balmer lines appear to re-turn to approximately their pre-outburst levels within∼ 30 days, but the He i lines appear to drop to a factorof 2–10 below this. This effect has been noted previouslyby Barsukova et al. (1998), Orlandini et al. (2000) andJaschek & Andrillat (2000) and is clearly real as can beseen in Fig. 1 where spectra before and after the outburstare compared. Further, there appear to be systematic dif-ferences between He i lines which are only manifested afterthe outburst: during outburst the He i EW ratios are sim-ilar to those before the outburst, but afterwards they candiffer, with He i 7065 A closest to the pre-outburst leveland He i 4713 A furthest below it. These differences arelarge, corresponding to a change by a factor of three inthe 4713 : 6678 ratio for example, so are not simply dueto changes in the continuum shape. This change in lineratios suggests changes in the physical conditions in theemitting gas; temperatures and/or densities; rather thanabundance changes (e.g. due to ejection of material assuggested by Orlandini et al. 2000). An increase in the6678 : 4471 ratio would be expected from a decrease intemperature, for example (e.g. Osterbrock 1989). Changesin the optical depth could also change ratios, and an in-crease in the 7065 : 4471 ratio could result from an increasein the optical depth (e.g. Osterbrock 1989). It is unfortu-nately impossible to be more quantitative for such opti-cally thick NLTE lines. Whatever changes are involved,they should leave the Balmer lines essentially unaffected.

We show a selection of H i, He i, and He ii line pro-files from outburst WHT/UES data in Fig. 6. Robinsonet al. (2002) have already discussed the profiles seen inoutburst. The WHT/UES data were obtained a few daysearlier than those of Robinson et al. (2002) but are quitesimilar. They argue that the line profiles, or at least thewings, are kinematic in origin and hence that emissionseen to ∼ 2500km s−1 indicates the velocity of the hy-drogen and helium emitting material outflowing from thesgB[e] star. There are alternative possibilities to be con-

Fig. 6. Profiles of the stronger hydrogen and helium lines inoutburst based on WHT/UES data from 1998 April 9. The pro-files have been rebinned by a factor of four for clarity. Regionscontaminated by other lines have been omitted.

sidered, however. The broad wings to the lines are mostprominent during outburst so may be associated with ma-terial ejected from the accreted compact object ratherthan from the sgB[e] star. Given that these lines are likelyformed in regions of high optical depth (see Sect. 7.4), in-coherent Thomson scattering may also broaden the linessignificantly; this has been proposed for other sgB[e] stars(Zickgraf et al. 1986). Scattering would be expected to besymmetric, but the observed profiles can be interpretedas the sum of a broad blue shifted component and a nar-rower component at rest (Robinson et al. 2002). Theseauthors obtained a satisfactory fit to the profiles witha double Gaussian, with the a narrow rest component(FWHM 50–85km s−1) and a broad component (FWHM∼ 160km s−1) blue-shifted by ∼ 60 km s−1. We find sim-ilar results from our WHT/UES spectra, although we donot obtain a satisfactory fit with a double Gaussian modelto the He i lines. This is likely due to the inadequacyof a Gaussian fit rather than an intrinsic asymmetry. Adouble Voigt profile fit, for example, gives extra freedomwhile still using symmetric components and works muchbetter. There is some sensitivity of the fit parameters towhether a Gaussian or Voigt profile is used, even whenboth fit well. For all fits to all lines we obtain blue shiftsof 40–110kms−1, and FWHM of 60–90km s−1 and 140–230 kms−1 for the narrow and broad components respec-tively, consistent with the estimates of Robinson et al.(2002). Only the blue shift, 40–110km s−1, needs to bekinematic and we then have much lower velocities in thehydrogen and helium emission regions, more comparableto the velocities inferred for metallic lines. It remains pos-sible that the width of the broad components is kinematichowever.

During the outburst, large changes are seen in theseline profiles. This is illustrated for Hα and He i 6678 Ain Fig. 7. At the earliest epoch (1998 April 3) both lines


are extremely broad and show structure not discernibleat later times. Emission is clearly seen up to 2500km s−1

in both lines, and possibly extends to ∼ 5000km s−1 inthe blue wing of Hα, although this is contaminated byother lines. The Hα profile in fact looks remarkably likethat of Hen S134, a near pole-on sgB[e] star in the LMC(Zickgraf et al. 1986). Both show a low-velocity blueshifted‘notch’, although the overall width of the Hα line is muchlarger for CI Cam. For Hen S134, Zickgraf et al. (1986)suggested that the notch was due to an unresolved ab-sorption feature. That might also be true for CI Cam, butthe He i profile suggests an alternative explanation. Thelatter shows inflections in both the blue and red wings. Infact this line clearly suggests two distinct components: anarrow line at rest and a broad blue-shifted line; the Hαprofile may arise the same way, but with a broader restcomponent. This is of course exactly the decompositionthat was applied successfully to later spectra and suggeststhat rather than using such a two-component fit as a con-venient parameterisation of an asymmetric profile, we canactually take it more literally: there really are two distinctemission regions. Furthermore, fitting the same two com-ponent model to the 1998 April 3 He i profile gives verysimilar results to the later spectra with a component sepa-ration of ∼ 50 kms−1 and a narrow component FWHM of∼ 75 km s−1. The main difference is that the broad com-ponent in the earlier spectrum is much broader with aFWHM ∼ 750km s−1. In fact, there is a hint that thenarrow component of Hα on April 3 may itself be morecomplex. The distortion of the peak may indicate that itcontains two unresolved components. This structure is re-peatable over all of the individual spectra, obtained onthis night, and the line peak was not saturated.

In quiescence, the weaker hydrogen and helium linesbecome narrower, approaching the width of the iron lines(see Figs. 7–9). These lines can never be expected to beas narrow as the iron lines as the thermal widths will belarger for lighter elements. Robinson et al. (2002) observedbroadening of 3.1 kms−1 for the Fe ii lines, and also esti-mated a temperature of the iron emission region of 8000K,corresponding to a thermal component of broadening of1.9 km s−1. The latter will correspond to ∼ 14 km s−1

for hydrogen lines. Fig. 9 shows the observed Hδ pro-file (which should be less affected by optical depth effectsthan stronger lines) from 2000 December 1, well after theH i lines had stabilised after the outburst. We have con-structed a simple model profile by taking a square toppedprofile extending to ±32 km s−1 (the model Robinson etal. 2002 used for the metallic profiles) and broadened itby ∼ 14 kms−1. This model profile is somewhat narrowerthan the observed Hδ profile, but the difference is not dra-matic, so it is plausible that the underlying H i kinematicsare similar to the Fe ii lines. Some extra broadening mayarise from a higher temperature in the hydrogen emissionregion and/or radiative transfer effects.

It appears that He ii 4686 A is still detectable after theoutburst, as noted by Barsukova et al. (2002). A close ex-amination of the pre-outburst OHP spectra reveals that

Fig. 7. Changes in the profiles of Hα and He i 6678 A fromoutburst to quiescence. There is clearly a dramatic change inthe width of the lines, and the early profiles also show a twocomponent structure.

it was also present then at a higher level, albeit still weak,indicating that it is not simply a remnant of the outburst.This is somewhat surprising and another contaminatingline cannot be ruled out. Such a line would, however, haveto share the property of the He i and N ii lines of beingweaker after the outburst than before, which is not con-sistent with Fe ii line behaviour, for example. At least oneother sgB[e] star shows He ii emission (the luminous B0star Hen S134; Zickgraf et al. 1986), however, so this isnot unprecedented, and is further evidence that CI Camis among the hottest sgB[e] stars.

7.2. The Bowen blend

A common feature of the spectra of X-ray binaries is theBowen blend, a mixture predominantly of N iii and C iiilines spanning ∼ 4635–4650 (McClintock, Canizares &


Fig. 8. Change in the width of the He i 5015 A line betweenoutburst and post-outburst phases. Note the similar behaviourof the N ii 5001 A line. The upper spectrum has been offset byone unit vertically.

Tarter 1975). This feature does also appear to be seen inCI Cam, but can only be discerned in the WHT/UES out-burst spectra. These data are shown in Fig. 10. Featuresare seen corresponding to first two of the N iii 4634, 4641,4642 A lines; the third may be present but unresolved from4641 A. These features appear broader than Fe ii and Cr iilines in the same region and are clearly declining muchfaster. It therefore seems likely that these are the N iiiBowen fluorescence lines. The adjacent echelle order alsoappears to show the C iii 4647, 4650, 4651 A lines whichshow similar properties. Neither set of lines appear tobe present after the outburst, although the post-outburstdata are of lower resolution and quality.

7.3. The 5650–5700 A blend

There are several lines between 5650 and 5700 A which aredramatically enhanced during the outburst. In many spec-tra these are blended with other lines in the region whichare less variable. The blend appears to be composed ofN ii, Sc ii, and Fe ii lines (Fig. 11); the variable ones arelikely allowed N ii lines from multiplet 73 at 5676, 5686 A.The strongest predicted line from this multiplet, 5679A,is not clearly seen, but may be present at a lower level andpoorly resolved from the wing of the 5676 A line. The evo-lution of the total equivalent width of the blend is plotted

Fig. 9. Hδ profile from 2000 December 1. The underlying ab-sorption (see Sect. 5) is much broader than this, so provides aneffectively flat background which has been subtracted. The his-togram is the data, the smooth line is a square profile extendingto ±32 kms−1 broadened by 14 kms−1. The model profile hasalso had instrumental broadening applied, although this is asmaller effect.

Fig. 10. N iii Bowen fluorescence lines seen in outburst. Threenormalised spectra are plotted corresponding to the threenights observed. The only major changes are in the Bowenfeatures which become systematically fainter with time. Thethree N iii components are indicated, although the 4642 A line,indicated by the dashed line, is not clearly visible and may beweaker and/or unresolved.

in Fig. 5d and shows a rather similar behaviour to the He ilines, with a lower flux after outburst than before.

7.4. Near-IR oxygen lines

There are a number of important emission lines of O i de-tected in the near-IR spectrum. These were not observedoften enough to construct lightcurves, but they still revealuseful information. The strongest is O i 8446 A. This hasan equivalent width of 41 A before the outburst and 57 A


Fig. 11. The 5650–5700 A blend. This appears to be a mix ofN ii, Sc ii and Fe ii lines. In the earliest outburst spectra theblend is unresolved, but it is clear that the N ii 5676 A linedominates, more so than in the spectra plotted here. The N ii5679 A line should also be present; it may be detected in thered wing of the 5676 A line. The upper spectrum has been offsetby 0.25 units vertically.

after it. It is very strong but as discussed by Grandi (1980)there are several possible reasons for this. Our observa-tions yield rather similar results to those of Chakrabarty& Roche (1997) for GX 1+4, so we come to similar con-clusions: that O i emission is produced by H i Lyβ fluores-cence; Clark et al. (1999) came to the same conclusion in-dependently from the 1.13 and 1.32µm infrared lines. Wefind that the O i 7774 A line is much weaker than 8446 A(EW ∼ 2.6 A before the outburst and 4.2 A afterwards),corresponding to dereddened flux ratios (7774:8446) afterthe outburst of order 0.07–0.10 accounting for uncertaintyin reddening. This rules out production of 8446 A by re-combination or collisional excitation, instead favouring afluorescence process. Continuum fluorescence would alsopredict O i 7254 A and 7990 A, neither of which are de-tectable before or after the outburst, arguing against thisinterpretation. The most likely explanation is then fluo-rescence by H i Lyβ. This process requires a gas which isoptically thick to Hα, and the ratio of 8446 A to Hα canbe used to estimate the Hα escape probability (see Grandi1980). We estimate a dereddened flux ratio (8446:Hα) af-ter the outburst of 0.07–0.12 and hence an Hα escapeprobability of (1.5 − 2.5) × 10−4, implying a very highoptical depth in Hα.

7.5. Metallic lines

CI Cam was dubbed ‘the iron star’ by Downes (1984) andthis is an appropriate designation. The spectrum before,during, and after the outburst is rich in Fe ii lines, in manycases blended in all but the highest resolution spectra. Itis impractical to show the evolution of all lines because ofboth the large number of lines and also blending problems.In Fig. 5e we compile results from four multiplets covering

a range of wavelengths and excitations. Lines from eachmultiplet that cannot be reliably deblended or that aresampled by few spectra are excluded. It is clear that thedecay from outburst is extended; it may still be continu-ing even in the most recent observations. This is consistentwith the observation that the iron lines appear stronger inpost-outburst spectra than in pre-outburst ones (if theyare still enhanced above their ‘true’ quiescent level). Minordifferences appear between multiplets, e.g. compare mul-tiplets 40 and 49. There is no systematic trend either withwavelength or with excitation, however, and the origin ofthe difference is unclear. The rise in the early stages ofthe outburst is real, as it involves a difference of a factorof two and is repeated across 12 lines for which measure-ments could be made over 1998 April 3–6. It suggests thatthe iron EW peaked around day 3–10 of the outburst.

Iron is by no means the only metallic species in thespectrum. For example, Robinson et al. (2002) also iden-tify permitted lines of Na i, Mg i, Si ii, Ca ii, Sc ii, Ti ii, andCr ii as well as C, N, and O lines. The behaviour of theselines can differ dramatically. For example Fig. 12 shows aselection of Fe ii and Ti ii lines during and after the out-burst. The Ti ii lines are clearly much more variable thanthe Fe ii lines and are almost undetectable in quiescence.This is puzzling as the Fe ii lines shown are of higher exci-tation than the Ti ii lines (5.6 eV for Fe ii; 4.0 eV for Ti ii.),and the ionisation potential of Fe ii, 7.9 eV, is also higherthan that of Ti ii, 6.8 eV. Possibly the Ti ii lines are formedin a different region to Fe ii. Cr ii lines may also virtuallydisappear in quiescence, although all are quite weak evenin outburst.

The profiles of the metallic lines have been charac-terised by Robinson et al. (2002) as a square topped profileextending to ±32km s−1 and subject to a small thermalbroadening. They attributed this to a uniformly expand-ing spherical shell (Fig. 13a). We will discuss these linesin the context of sgB[e] models in Sect. 9.1, but here willnote a complication and an alternative explanation. Thecomplication is that while a square topped profile clearlyprovided a good description of the profiles at the time ofthe Robinson et al. (2002) observations, this is not ade-quate at other times. Fig. 12 shows two segments of our1998 April 9 WHT/UES outburst spectrum, together withpost-outburst counterparts. The latter are at lower reso-lution, but are still sufficient to show deviations from theflat topped profile; the Fe ii 4297 A line clearly shows dou-ble peaks, or a central depression. This is essentially thesame as the [Fe ii] 4287 A line, but the latter is also dou-ble peaked in outburst. Other lines in this spectrum showsimilar profiles, so this is not just due to noise. The prob-lem is worse than this, however, as can be seen in Fig. 12b.This shows our WHT/UES spectrum of the region usedfor Fig. 7 of Robinson et al. (2002). Two lines in partic-ular, Cr ii 5421 A and Fe i 5430 A, both of which showedflat topped profiles in the data of Robinson et al. (2002)are actually double peaked at our earlier epoch. Thus thedeviation from a flat topped profile cannot simply be apost-outburst effect, and the profiles appear to evolve from


Fig. 12. a) Changes in metallic lines from outburst to quies-cence. Ti ii lines are much more variable than Fe ii lines and arevirtually undetectable in quiescence. The Fe ii profiles change,being rounded in outburst but concave in quiescence (like [Fe ii]lines). Note that the Fe ii line at 4297 A in particular does notbecome square topped in quiescence. The upper spectrum hasbeen offset by one unit vertically. b) Outburst spectrum of5415–5437 A region. The spectrum from April 14–17 is repro-duced from Fig. 7 of Robinson et al. (2002). Note how theconcave tops of the Cr ii and Fe i lines in the earliest spectrumlater become flat tops

a double peaked to flat to double peaked form throughand after the outburst. Consequently the flat topped pro-files seem to be the exception rather than the norm, andthe variations seen suggest some deviation from sphericalsymmetry. We suggest that the underlying symmetry isaxial rather than spherical, but that our viewing angle isalong the axis (i.e. we see the star pole-on). If we viewany equatorial section of a spherical outflow pole-on thenthe profile will be rectangular, with narrower equatorialoutflows producing narrower profiles (Fig. 13b). If emis-sion from the equatorial outflow varies with latitude thenvariations from a flat top would be seen, and a centraldepression would imply less emission near the equatorialplane (Fig. 13c).

+v−v

−v +v

+v−v

(c)

(a)Flux

(b)

θ

Fig. 13. Illustrative optically thin profiles for different outflowgeometries. The viewing angle is taken to be from the top of thepage, i.e. pole-on. a) A spherically symmetric outflow producesa square profile extending to ±Vterm. b) An equatorial sectionof a spherical outflow also produces a square profile (whenviewed pole-on), but narrower, extending to ±Vterm sin θ. c) Ifthe outflow is non uniform, with less emission closer to theequatorial plane, then a central depression is created, becausethere is less low velocity emission.

7.6. Forbidden lines

A number of forbidden transitions are observed in [O i],[O iii], [N ii] and [Fe ii]. These are a rather heterogeneousselection of lines spanning a range of ionisation stagesand showing very diverse behaviours. As already noted byRobinson et al. (2002), [N ii] and [O iii] lines are extremelynarrow, more so than any permitted lines (e.g. Fig. 8).[O i] and [Fe ii] lines, however, exhibit square-edged pro-files similar to the Fe ii lines, although typically with amore pronounced central depression (e.g. Fig. 12a). In thecontext of the equatorial outflow geometry suggested forthe Fe ii lines, a deeper central depression would imply lessemission close to the equatorial plane. This makes sense,as the density is expected to be higher there and so for-bidden lines may well only be seen from higher altitude,lower density, layers.

The line changes with time differ between speciestoo. Unfortunately, many of the lines are weak and/orblended with other lines, principally of Fe ii. The onlyline well suited to quantitative analysis is [N ii] 5755 A,as it is strong and unblended. The equivalent width evo-lution of this line is shown in Fig. 5f. The early evolu-tion (<∼ 10 days) is roughly consistent with a constant lineflux, with the equivalent width increasing as the contin-


uum fades. Comparison with outburst photometry (Clarket al. 2000), after correcting for the increased contribu-tion of the lines to broad band flux (Section 6), indicatesthat this outburst line flux is comparable to that seen in1998 January. Barsukova et al. (2002) note that the lineflux does increase modestly around 50–250days after out-burst, then decays again. The decay can clearly be seenin Fig. 5f, although the rise is hard to distinguish fromthat due to the decreasing continuum flux without fluxcalibrated spectra.

The [O iii] 5007 A line also exhibits a narrow profilebut is clearly stronger in outburst than after it (Fig. 8).The 4959 A line behaves in a similar way. These lines aredifficult to measure in most spectra as they are weak andblended, but it is clear that the outburst flux is muchgreater than pre-outburst and that after the outburst theydecline, but like [N ii], they do not appear to have droppedto the pre-outburst level. For example, for [O iii] 5007 A,the pre-outburst EW is difficult to measure but ∼ 0.25 A.On 1998 April 11 it had an EW of 2.0 A and by 1998 July20 this had dropped to 1.1 A. As late as 2001 October 23it was still >∼ 0.8 A. The decline in EW from 2000 Aprilto July corresponds to a decrease of about a factor ofthree in line flux. Since the [O iii] lines have very long re-combination times, this implies that significant collisionalde-excitation must be occurring, possibly when the radioejecta reach the line formation region. Assuming an ex-pansion velocity of ∼ 5000km s−1, then a reduction of afactor of three within ∼ 100days implies that most of the[O iii] emission originates within ∼ 300AU.

[Fe ii] lines also strengthen moderately during outburst(e.g. Fig. 12a) and [O i] lines increase enormously, beingalmost undetectable before and after the outburst.

8. Short term spectral variability

There have been some claims of short term variabilityin CI Cam during outburst. At X-ray energies Fronteraet al. (1998) found with Beppo-SAX that the 0.5–1.0 keV lightcurve showed significant variations on ∼ 100 stimescales on 1998 April 9–10. They found no significantvariation from a smooth decay in the simultaneous 1.5–10 keV data however, nor were any variations seen on April3. Ueda et al. (1998) examined ASCA data from 1998April 3–4. They also found that there was no variabilityabove 1 keV, but that soft flares were seen below 1 keV ontimescales of a few hours. Belloni et al. (1999) examinedRXTE data spanning April 1–9 and found no evidencefor any variation other than a smooth decay at any time.The short term X-ray variability thus appears confined toflares in the soft (<∼ 1 keV) band.

Optical variability appears sporadic too. Frontera etal. (1998) also found evidence for 0.3mag optical flickeringon hour timescales on April 6, but not in later observa-tions spanning April 10–26. Clark et al. (2000) examinedphotometry from April 13 and 19 and found no variationswith amplitude greater than 1 percent.

Fig. 14. Change in the strength of the He i 6678 A line inMcDonald spectra from 1998 April 20. Spectra were normalisedto a flat continuum which was then subtracted. Integrated linefluxes are plotted as fractional variations about the mean.Error bars are formal errors obtained in extracting one-dimensional spectra and do not include additional uncertain-ties that may be introduced by normalisation of the spectra.

The McDonald spectra obtained during the declinefrom outburst were taken as a series of short exposuresto facilitate studies of line variability. Because a relativelynarrow slit was used, however, there are significant slitlosses. Fortunately, as noted above (Clark et al. 2000),there was no detectable short-term variability in simulta-neous R band photometry, so we chose to normalise eachspectrum before extracting emission line lightcurves; theselightcurves are effectively equivalent widths. We find noevidence for variability on timescales from 3min to ∼ 2 hrin Hα or Fe ii lines. Formally we find rms scatters in thelightcurves of 1.1–1.6 percent in Hα and 0.8–1.5 percent fora composite of several iron lines. The one line that may ex-hibit short term variability is He i 6678 A. This shows 1.9,4.9 and 3.5 percent rms variations on 1998 April 18, 19 and20 respectively. On the latter two nights there is clearlya systematic variation. To be sure this is not an artifactwe have renormalised the continuum using just the 6648–6663 and 6693–6708A regions and recentered the line ineach spectrum by cross-correlating the line profiles withan average. None of these changes affect the result: thisline appears to show real variability. The case is most per-suasive on the last night. There is a clear overall rise in thelinestrength of ∼ 10percent over < 3 hours, shown in Fig.14. This is in the opposite direction to the overall declinein the linestrength and much larger than expected fromthe changes in the continuum strength alone (e-foldingdecay time ∼ 24 days at this time; Clark et al. 2000). Wehave also constructed rms spectra for each night and com-pared them with the average spectra. These suggest thesame conclusion as the lightcurves; the only feature thatseems to show significant variability is the He i line on thesecond and third nights.


We also examined the FLWO data from 1998 April3 in the same way. The rms spectrum shows no variablefeatures other than Telluric bands. The line strengths forHα and He i 6678 A show an rms scatter of 1.8 percentand 1.2 percent respectively with no systematic trend. Webelieve this represents a null detection of variability onthis night.

9. Discussion

9.1. CI Cam as an sgB[e] star

It is clear that the X-ray outburst was associated with dra-matic changes in the spectrum of CI Cam. Having iden-tified it as an sgB[e] star, however, it will prove useful tofirst compare it with the small number of other stars ofthis class and see how much of its ‘unusual’ behaviour,is typical of these objects, and what requires additionalinput from a compact object.

Robinson et al. (2002) infer a predominantly sphericalgeometry for the outflow from CI Cam. Several compo-nents are suggested. A cool, low velocity (32 km s−1) windgives rise to the many iron lines and the square profilesof these lines suggested that this component is spherical.Multiple higher velocity components (> 1000km s−1) aresuggested by hydrogen and helium lines, and by the UVresonance lines. This picture has several problems as notedby Robinson et al. (2002), principally i) it is unclear howseveral near-spherical winds can co-exist with very differ-ent velocities and ionisations; and ii) the quiescent X-rayluminosity is very low for a compact object continuallyburrowing through a dense, spherical outflow.

This picture is also rather at odds with existing mod-els of sgB[e] stars (Zickgraf et al. 1985,1986; Oudmaijeret al. 1998). The structure inferred from other systems isof a two-component wind which is far from spherical. Ararefied, hot, high velocity wind dominates in the polarregions and is responsible for the UV lines often seen withP Cygni profiles. This wind is identical to the winds of nor-mal early-type supergiants. Other lines are attributed to acooler, denser wind concentrated in the equatorial regionsthat is not present in normal supergiants. The nature andorigin of this equatorial material remain unclear, and it isprobably quite different to the discs around classical Bestars. Can CI Cam be fitted into this framework?

The interpretation of the UV lines seen in CI Camis most straightforward. These show broad P Cygni linestypical of supergiant winds. They can be associated withthe hot, high velocity, polar outflow, and we essentiallyagree with Robinson et al. (2002) on the interpretation ofthese lines.

The hydrogen and helium lines are more problem-atic, however. In outburst, these lines are broad andasymmetric, with blue-shifted emission extending to >∼2500km s−1, but with no detectable absorption compo-nents. Robinson et al. (2002) associate these lines with ahigh-velocity, weakly collimated outflow from the sgB[e]star. It is unclear where this could be situated, however:

if it originated from the same region responsible for theUV lines we might expect P Cygni line profiles, as seen inHD 87643 (Oudmaijer et al. 1998). The latter object maybe rather different to other sgB[e] stars with a higher H icolumn density in the polar wind. In most sgB[e] stars,hydrogen and helium emission is instead attributed to theequatorial component, and hence relatively low velocitiesshould be involved, as is seen in the iron lines. In fact,a clue that the hydrogen, helium and iron lines are allassociated with the same region, or regions, is providedby similarities of the line profiles. Robinson et al. (2002)associate Fe ii lines with a very different region to thehydrogen and helium lines, but they also note that ‘Thestronger [iron] lines have more rounded profiles, becomingalmost Gaussian in shape as the lines become stronger,and the very strongest lines have an extended blue wing.’Another way to put this is that the strongest lines be-come more akin to hydrogen and helium lines. While therounded profiles could indicate radiative transfer effects,the presence of asymmetry in the Fe ii lines suggests thatthe same rest and blue-shifted components are present inthe iron lines as are seen in hydrogen and helium pro-files. If the strongest, optically thickest iron lines look likehydrogen lines, then the corollary is that the weakest, op-tically thinnest hydrogen and helium lines should look likeiron lines. This does seem to be the case, as demonstratedin Sect. 7.1 and Fig. 9.

The decomposition of the hydrogen and helium linesinto two clear components suggests that two distinct re-gions are involved during outburst. The narrow rest com-ponent dominates in the iron lines and is also importantin the hydrogen and helium lines. A second broad, blue-shifted component is present in hydrogen and helium lines,and weakly detectable in the strongest iron lines. The nar-rower, rest component is likely the same component asseen in quiescence, and in other sgB[e] stars, originatingfrom the equatorial outflow. The profiles seen in H i andHe i are very different to the Fe ii lines, but the formerare broadened by larger thermal widths and incoherentThomson scattering, as suggested by Zickgraf et al. (1986)for other sgB[e] stars. The strong, broad, blue-shifted com-ponent is a feature of the outburst and may not be associ-ated with the outflow from the sgB[e] star at all. We willdiscuss this possibility further in the following section.

Associating the iron lines with the equatorial materialconflicts with the interpretation of their profiles offered byRobinson et al. (2002). They argue that the rectangularprofiles allow little deviation from spherical symmetry andno significant rotational velocity. As we have discussed inSect. 7.5, however, large deviations from spherical geom-etry are possible in one case: that an equatorial outflowis viewed pole-on. A pole-on equatorial outflow can alsoexplain the apparent lack of rotational velocities, as theprojected rotational velocity could be very low, and theintrinsic rotational velocity in regions of the disc whereFe ii can arise is in any case likely to be very low.

As a further argument for the presence of a sphericallow velocity wind, Robinson et al. (2002) claim on the ba-


sis of the scaling relation of Bjorkman (1998) that dustcan condense in this wind and that an equatorial outflowis therefore not required for the production of dust. Thescaling relation of Bjorkman (1998) is based on the ar-gument that the density of wind compressed outflows insgB[e] stars at radii where the wind is cool enough fordust to condense is ∼equal to those of post-AGB stars(which do form dust) and therefore that it is possible fordust to condense in such outflows. However, the scalingrelation does not allow for the fact that the winds in post-AGB stars are significantly enhanced in species such asSi and C (via dredge up) which form dust up to a fac-tor of ∼10 times more easily relative to the winds fromsgB[e] stars. Additionally, the mass loss for the sgB[e] staris calculated from the base density of the outflow at thesurface of the star and the terminal velocity of the outflow,rather than the velocity of the outflow at the surface ofthe star. This will also serve to overestimate the density ofthe outflow at the dust condensation radius. Combiningthe 2 arguments we estimate that the density is likely tobe 2–3 orders of magnitude too small to allow dust tocondense (J. Porter, priv. comm.). While this does not apriori argue against the presence of a quasi-spherical windresulting in the Fe ii line profiles it does demonstrate thatdust cannot condense in it and that a region of higherdensity – whether an equatorial outflow or dense ejecta– is required for condensation. Robinson et al. (2002) ad-ditionally argue that the presence of a resolved, circulardust shell (Traub, Millan-Gabet & Garcia 1998) is consis-tent with their interpretation, but we note that if a disc isviewed pole-on then dust in the outer regions of the discwill produce the same apparent structure.

Forbidden lines are a rather heterogeneous mixture,but broadly divide into two types. Narrow lines ([N ii],[O iii]) are formed in a region of very low bulk, turbulentand/or thermal velocities. This probably places them out-side the two component outflow from the sgB[e] star, andthey may be a remnant of an earlier phase of evolutionof one of the stars (c.f. Oudmaijer et al. 1998). The [Fe ii]lines show profiles similar to the Fe ii lines, but with a morepronounced central depression. As discussed in Sect. 7.6,this suggests that these lines may be formed in the upperlayers of the same equatorial material responsible for Fe ii;[O i] lines are likely formed in the same region.

To summarise, by analogy to other known sgB[e] starswe suggest that when the system is not in outburst,the strongest optical permitted lines (H i, He i, Fe ii, andothers) and some forbidden lines ([Fe ii], [O i]) arise inan equatorially concentrated region, whether a quasi-Keplerian disc (e.g. Okazaki 2001), a hybrid disc witha slow radial expansion velocity, or even in a radiationdriven disc wind such as is suggested in the Galactic sgB[e]star HD 87643 (Oudmaijer et al. 1998). UV P Cygni linesoriginate from a hotter, higher velocity polar outflow ofmuch lower density. The whole structure is viewed closeto pole-on, hence disc lines have rather low velocities asboth rotational and expansion velocities would be mainlydirected in the equatorial plane. Finally the narrow forbid-

den lines ([N ii], [O iii]) come from a much more extendedand near stationary region.

Our suggested model does depend on viewing thesource near pole-on. There are several independent lines ofevidence to support this, suggesting that CI Cam is seenat a lower inclination than many sgB[e] stars.

1. Most sgB[e] stars show low velocity blueshifted ab-sorption in their Balmer line profiles (Zickgraf et al.1985,1986; Oudmaijer et al. 1998). This can be inter-preted as due to absorption by the slow equatorial windcomponent. No such absorption is seen in CI Cam sug-gesting that the inclination is low enough that our lineof sight does not pass through the slow wind.

2. sgB[e] stars are expected to show significant po-larisation changes through Hα. The observations ofHD 87643 are qualitatively similar to what is expectedfor a rotating, expanding disc (Oudmaijer et al. 1998).In CI Cam no polarisation changes are seen throughthe lines (Ikeda et al. 2000). In the models of Wood etal. (1993) the polarisation decreases with inclination,so a lack of line polarisation also supports a low inclina-tion. In addition, the outburst observations of Ikeda etal. (2000) showed a continuum polarisation consistentwith surrounding field stars, and hence with purely in-terstellar polarisation. This is also as expected for anearly pole-on disc.

3. We have argued above that dust cannot form in aspherical shell and that it must be associated with theouter disc. The near perfectly circular resolved infraredimage then suggests that this disc must be viewed pole-on.

It therefore seems plausible that CI Cam is a pole-onsgB[e] star. Such a model is consistent with the propertiesof both the optical and UV spectra.

9.2. The nature of the outburst

We now move on to discuss the role of the compact ob-ject. Its nature obviously has some bearing on this discus-sion. The X-ray outburst was very short, much shorterthan typical of soft X-ray transients (SXTs). This, to-gether with the radio emission from an expanding rem-nant (Mioduszewski et al. 2002, in preparation) suggestssome kind of explosive event. Orlandini et al. (2000) sug-gest a thermonuclear runaway on the surface of a whitedwarf is responsible, but this seems hard to reconcile withthe high inferred X-ray luminosity (Robinson et al. 2002)and the γ-ray detection by CGRO/BATSE (Belloni et al.1999). It seems more likely that the outburst involved abrief burst of supercritical accretion onto a neutron staror black hole, resulting in ejection of much of the accretedmaterial. This accounts for the observed radio ejecta andpossibly the broad components of the hydrogen and he-lium lines. A supercritical accretion model could also in-volve a large enough optical depth of scattering materialaround the X-ray source to smear out any short timescalevariability.


One possible cause for such a large burst of accre-tion would be the passage of the compact object throughthe equatorial plane. We can attempt to estimate the ac-cretion rate, although this will be very uncertain as theorbital parameters are unknown and the physical con-ditions in sgB[e] star outflows are not well understood.We assume stellar parameters typical of hot, luminoussgB[e] stars (Zickgraf et al. 1986) and adopt a representa-tive 30 year orbit of a 10M� black hole around a 60M�sgB[e] star with an eccentricity of e ∼ 0.9. The perias-tron distance is then about 10–20Rstar and the perias-tron velocity 170km s−1. In the model of Oudmaijer etal. (1998) for HD 87643, the equatorial density at 10Rstar

is ∼ 2.5 × 10−12 g cm−3. This model may not be correct,and CI Cam is probably more massive and luminous thanHD 87643, but this at least indicates the kind of equa-torial density which is considered plausible. The velocityof the compact object relative to the equatorial materialwill be dominated by its orbital motion, so the Bondi ac-cretion rate is then predicted to be ∼ 6 × 1021 g s−1 or∼ 400 MEdd (Bondi 1952). Given the uncertainties men-tioned above, this is a very approximate estimate, andcould be at least an order of magnitude off. Even allowingfor this large uncertainty, the accretion rate in this sce-nario can therefore be expected to be extremely high, dueto the high density and low velocity of the equatorial mate-rial, and hence a supercritical accretion episode is possible.In contrast, when the compact object is out of the planethe density is lower by a factor of 104 (in the model ofOudmaijer et al. 1998) and the relative velocity is higher,since the high latitude outflow moves much faster. TheBondi accretion rate is then expected to be much lower,∼ 8×10−5 MEdd. If the accretion efficiency remained high,η ∼ 0.1, then CI Cam should still be a relatively bright X-ray source, with LX ∼ 1035 erg s−1. However, at these lowaccretion rates then the flow could be expected to becomeadvective as proposed for other quiescent black hole candi-dates (Narayan, Garcia & McClintock 2001 and referencestherein), and the accretion efficiency would then be lower,η ∼ 10−2–10−4, implying LX ∼ 1032–1034 erg s−1 as ob-served (Robinson et al. 2002). This model therefore avoidsthe problem that Robinson et al. (2002) had, that a com-pact object burrowing through a dense spherical outflowshould be persistently bright.

The passage of a compact object through the equato-rial material, and the ejection of a lot of material fromnear it could affect the equatorial region significantly, sochanges in the line spectrum from this region would beexpected. We cannot offer an exact mechanism for this,however, and several factors may be involved: the tidaleffect of the compact object passage; X-ray heating; andthe interaction of the expanding radio remnant with theequatorial flow. The rapid response of the lines to the out-burst, peaking within a few days of the X-rays, constrainsthe tidal effect, as a tidally triggered disc outburst wouldbe expected to proceed on a viscous timescale, if the disc isKeplerian (i.e. thin and rotationally supported, with smallinflow or outflow velocity). This is likely to be very long,

hundreds to thousands of days. It is far from clear thatsgB[e] discs are Keplerian, but other timescales, e.g. theoutflow time from the stellar surface to ∼ 50 stellar radii,are likely to be similar. The extended decay of the ironlines may indeed be due to a recovery of the disc to itspre-outburst state on this timescale, but the rise of theoutburst cannot be, so a tidally triggered disc outburst isunlikely.

The X-ray heating effect will obviously be much faster,and should rise and fade on a timescale comparable to theX-ray outburst itself, although could be prolonged by sig-nificant cooling or recombination times. This is not seen,although a signature of X-ray heating of the disc may bepresent in the form of strong 6–7 keV line emission seenby SAX (Frontera et al. 1998), ASCA (Ueda et al. 1998)and RXTE (Revnivtsev et al. 1999; Belloni et al. 1999).If this line is attributed to fluorescent iron emission thena large, cold, optically thick surface is needed coveringabout half the sky as seen from the X-ray source (Uedaet al. 1998). This is what would be expected if the X-raysource were above the plane of the equatorial material.The dramatic enhancement of high excitation lines, suchas He ii and N iii during outburst also suggests that X-rayor extreme-UV irradiation may place a role.

The expansion of the radio ejecta can more naturallyproduce the timescales observed than either a tidal inter-action or X-ray irradiation can. Assuming an expansionat ∼ 5000km s−1 (from Mioduszewski et al. in prepara-tion, assuming d ∼ 5 kpc), this would reach a disc radiusof ∼ 50 stellar radii (∼ 2000R�) in about 3 days, so this isconsistent with the relatively short outburst and the risetime of the Fe ii EWs. The expanding ejecta may also ex-plain the broad blue-shifted component seen prominentlyin hydrogen and helium profiles during outburst, as theinferred expansion velocities are comparable to the max-imum detectable velocities in the emission lines. This isan appealing interpretation, as the origin of these ejectais most likely the compact object, and hence an overallblue shift can be explained as the orbital velocity of thecompact object at the time of ejection. The overall blue-shift seen, 50–100km s−1, is reasonable compared to ourestimate of the periastron velocity ∼ 170 km s−1. It is alsosensible that we see a blue-shift, rather than red, as X-ray spectra taken in the outburst decay show a strongfluorescent iron line, consistent with reflection from theequatorial material, and do not show strong absorption;both indicate that the compact object was likely betweenus and the equatorial plane at the time, and hence shouldhave been moving toward us.

9.3. The origin of the optical continuum

The dramatic increase in the optical luminosity of CI Camduring the outburst is also intriguing. The optical rise wasnot observed, but if it was later than the X-rays then therise time was fast, <∼ 2 days. While large optical brighten-ing is typical of SXTs, it is actually rather surprising in a


system containing a hot, luminous sgB[e] star. The bright-ening is such that whatever provides the additional sourceof light during outburst must significantly outshine thesgB[e] star itself. Relative to the mean pre-outburst level(Bergner et al. 1995) the brightest magnitudes reportedare brighter by 1.9mag (U , Apr 3.9; Hynes et al. 1998),2.1mag (B, Apr 3.1; Garcia et al. 1998), 2.3mag (V , Apr3.1; Garcia et al. 1998), 3.5mag (R, Apr 2.1; Robinsonet al. 1998), 2.4mag(I, Apr 3.8; Clark et al. 2000). Theseobservations are not simultaneous and are not intendedto indicate the spectrum; merely to indicate that the out-burst was dramatic throughout the optical region. Sincethe peak of the outburst was missed in the optical theseare actually lower limits on the outburst amplitude. Whilesome of the brightening is due to stronger line emission,not all can be; removing the line emission from our earlyspectra only increases B and V by ∼ 0.6mag, so muchof the brightening has to be enhanced continuum. Fig. 4indicates that on 1998 April 3, the excess continuum was2–5× brighter than the sgB[e] star, and redder than it.Where can this extra continuum come from? Some possi-bilities which we discuss below are:

1. Heating of the sgB[e] star during the outburst.2. Heating of the dusty envelope around the system.3. Emission from an accretion disc around the compact

object.4. Enhanced emission from the circumstellar material

around the sgB[e] star.5. Emission related to the observed radio ejection.

To help constrain the possibilities it is useful to ex-amine the spectrum of the excess light. Fig. 15 shows thedifference between the early outburst and post-outburstspectra from Fig. 4. Close examination of the outburstspectrum, and comparison with high resolution spectrataken later in outburst, indicates that although much ofthe spectrum is dominated by emission lines, the lowestpoints do appear to indicate the continuum. The differ-ence spectrum is obviously rather uncertain, but manypossibilities can be ruled out. If it is a power-law, thenit is relatively flat in fν . The power-law model plotted isfν ∝ ν1/2. A completely flat spectrum would also be pos-sible, but it could not be much redder or bluer than thesepossibilities. A hot black body is ruled out, as indicatedby the fact that the difference spectrum is redder thanthe sgB[e] spectrum. A 10 000K black body is shown; atemperature much cooler than this would also not be con-sistent with the data. Similar conclusions can be drawnby examining the change in B − V during the outburst,from ∼ 0.8mag to <∼ 1.2mag. It is obviously difficult tobe precise given the uncertain reddening and large linecontribution, but a very hot or very cold black body, or avery red or very blue power-law can be ruled out.

This colour information can immediately be used toconstrain the origin of the excess light. Interpretation 1can be ruled out as heating of the sgB[e] star could notmake the spectrum redder even if a 5× increase in bright-

ness were possible. In this case, the excess should be ex-tremely blue. Equally, interpretation 2 can be ruled outas the excess light is not cool enough. Heated dust woulddissociate for T >∼ 2000K, but the observed spectrum andchange in B−V are clearly not consistent with such a lowtemperature.

The brightness of the outburst argues against interpre-tation 3. After correcting for line emission (∆V ∼ 0.6mag)and extinction (AV ∼ 4 mag), the peak dereddened con-tinuum magnitude is V <∼ 5.8 and the average quiescentvalue is 〈V 〉 ∼ 7.7mag; hence the additional continuumsource has V <∼ 6.0. At a distance of ∼ 5 kpc this corre-sponds to an absolute magnitude of MV <∼ −7.5. This ismuch brighter than any low mass X-ray binary (LMXB)(−5 < MV < +5; van Paradijs 1994), i.e. much brighterthan other observed accretion discs around black holesor neutron stars. If the additional light were an accre-tion disc it would then have to be more luminous thanthose in LMXBs. Since the spectrum indicates that it isnot extremely hot, it would have to be large. As the sizeof LMXB accretion discs is limited by tidal truncation,a much larger disc in CI Cam is certainly possible. If thedensity in the disc were comparable to the disc densities inLMXBs, however, such a large disc would provide enoughmass to sustain a much longer X-ray outburst than seen,and a decay timescale of months, more like that seen inSXTs, would be expected. Indeed Robinson et al. (2002)have argued that the very rapid decay timescale in CI Camindicates that the accretion disc, if present, must be verysmall. This is a very different environment from that in anLMXB, however, and the disc could be stabilised againstthermal instability by irradiation from the supergiant atmuch lower densities than required in an LMXB, resultingin a lower disc mass. The brightness remains a problem,however, as the disc luminosity would be comparable tothat of the supergiant and greater than that of the X-ray source at the same time, and so reprocessing of lightfrom either is unlikely to dominate. It therefore seems un-likely that an accretion disc around the compact objectcan dominate the optical brightening in outburst.

The relatively low temperature and large area inferredfor the excess light, if it originates in black body emission,can more plausibly be associated with the equatorial ma-terial around the sgB[e] star than an accretion disc aroundthe compact object. For a temperature of ∼ 10 000K,the extra emission requires an area ∼ 40× that of thesgB[e] star. Both the temperature and size are plausiblefor the equatorial material around a sgB[e] star. In thiscase the timescale arguments are similar to those for thelines. The optical decay appears slower than the X-rays(e.g. Clark et al. 2000). The peak optical luminosity ofthe continuum source is also large; for T ∼ 10 000K andan area ∼ 40× that of the star, the luminosity was at least∼ 2×105 L� ∼ 8×1038 erg s−1 (assuming d ∼ 5 kpc). Thisis actually larger than the peak X-ray luminosity at thisdistance, LX ∼ 3 × 1038 erg s−1, and by the time of theoptical observations, the X-ray flux had already droppedby more than an order of magnitude (Belloni et al. 1999).


Both the decay timescale and optical luminosity render itunlikely that the continuum brightening can come fromX-ray reprocessing. The timescales would be more plausi-ble for an interaction of the equatorial material with theradio ejecta; assuming an expansion speed ∼ 5000km s−1,this area will be covered in less than half a day, so a rapidoptical brightening is possible in this way, i.e. powered bythe kinetic energy of the ejecta.

The remaining possibility is that there is a consider-able direct optical contribution from the ejecta from theexplosion. The similarity of timescales of the optical andradio decay (Clark et al. 2000) argue for this interpre-tation. Optical synchrotron can probably be ruled out.The dereddened magnitude of the additional continuumcomponent of V <∼ 6 corresponds to a flux of >∼ 15 Jyearly in the outburst. Contemporaneous radio observa-tions were much weaker than this, with the spectral energydistribution peaking at 650mJy at 8.4GHz and decreasingat higher frequencies (Hjellming & Mioduszewski 1998b).Optical synchrotron emission thus seems unlikely withouta very unusual electron energy distribution. However, ifthe accretion rate was highly supercritical, we might ex-pect much of the energy released to be reprocessed bya scattering envelope to produce a bright optical source(Shakura & Sunyaev 1973). Such a supercritical accretionscenario has been invoked for a number of X-ray binaries.SS 433 is likely to a persistently supercritical source (e.g.Fabrika 1997 and other authors; see also Okuda & Fujita2000 and references therein). V4641 Sgr, which also hadan extremely short but luminous X-ray outburst has beensuggested to have undergone a transient burst of supercrit-ical accretion (Revnivtsev et al. 2002). Shakura & Sunyaev(1973) predict that the optical spectrum in the supercriti-cal regime should appear as a power-law, Fν ∝ ν1/2, satu-rated with broad emission lines. We have already discussedthe origin of the broad emission lines that we see and sug-gested that these may be related to outflowing materialfrom supercritical accretion. The outburst optical contin-uum emission is clearly redder than the underlying star,and the excess continuum emission is consistent with aFν ∝ ν1/2 power law (Fig. 15). This therefore does seema plausible interpretation. If this supercritical accretionscenario is correct, then the observations suggest that theaccretion rate was extremely high and hence the spheri-sation radius large; assuming a terminal velocity for theejected material of ∼ 5000km s−1, and an optical luminos-ity ∼ 1038 erg s−1, suggests an accretion rate of M ∼ 102–104 Mcrit (Shakura & Sunyaev 1973), consistent with ourestimate above of the possible periastron accretion rate.

All of these calculations are extremely simplistic. Ouraim is not to present a detailed spectral model, but ratherto test which explanations of the outburst emission areplausible. X-ray heating, whether of the sgB[e] star it-self, the equatorial material, an accretion disc around thecompact object or extended dust, is not consistent withobservations. Heating of the equatorial material by theinteraction with the expanding radio remnant, or directemission from these ejecta, remain possibilities. The lat-

Fig. 15. Spectrum of the extra light seen in outburst, obtainedby taking the difference between spectra obtained on 1998April 3 and 1998 October 29. This has been dereddened us-ing the Fitzpatrick (1999) extinction curve assuming AV = 4.0.Two possible models for the continuum spectrum are also plot-ted; the power law corresponds to Fν ∝ ν1/2.

ter has the advantage that invoking a supercritical accre-tion regime explains both the optical continuum emissionand the broad, blue-shifted emission components, so is ourfavoured interpretation, but it is likely that a combinationof these mechanisms is involved.

9.4. Is CI Cam an X-ray binary?

Finally we emphasise that none of the preceding discus-sion depends on CI Cam being a binary system. We, andothers, have argued that CI Cam is an sgB[e] star, and acompact object of some kind is clearly implicated in theX-ray outburst, but it is not necessary for them to bephysically associated. Indeed, the properties of CI Camin quiescence are sufficiently similar to other hot, lumi-nous sgB[e] stars that if it is a binary then the compactobject’s influence appears to only be important during anoutburst. For example, it clearly does not truncate thedisc as can happen in classical Be X-ray binaries. Whileit could be that CI Cam does have a compact object ina long period orbit, it is also possible that the outburstcould have resulted from a chance encounter of an isolatedblack hole or neutron star with the circumstellar material.As a massive, young object, CI Cam is likely located ina region of recent star formation, so there are likely tobe many stellar remnants in its proximity. While such achance encounter still seems improbable, it cannot be dis-proved based on the statistics of a single event. So far wehave no conclusive evidence for the binarity of CI Cam.Barsukova et al. (2002) have suggested that two periodic-ities are present. Their short 11.7 day period is certainlytoo small to represent the orbital period in a system suchas this as the compact object would have to be almost onthe surface of the sgB[e] star, or even inside it. The longer1100day period would be more plausible, but cannot be


considered convincing until it is seen to repeat for multiplecycles. In balance, while the hypothesis that CI Cam is abinary seems most likely, it is not proven and the alterna-tive, that there was a chance encounter with an isolatedcompact object, cannot be ruled out.

10. Summary and conclusions

CI Cam is an sgB[e] star, and many of its characteristics,particularly the emission line spectrum, are representativeof this class. Comparison with other sgB[e] stars suggeststhat CI Cam is among the hottest members of the classand viewed nearly pole-on. It differs from the other sgB[e]stars in interacting with a compact star which introducesan additional variable element. It seems most likely thatthe compact object is physically associated with CI Cam,i.e. that it is an HMXB. In this case, the compact ob-ject is probably in a long period, eccentric orbit. Howeverwe cannot rule out the possibility that this was a chanceencounter and that the two stars are not physically asso-ciated. Resolution of this issue will likely involve waitingfor a true periodicity, repeated over several cycles, or an-other outburst. In considering the outburst mechanism,there is actually little difference between a long period ec-centric orbit and a chance encounter, so our discussion ofthe outburst applies to both cases.

We suggest that the majority of the optical emissionlines originate from an equatorially concentrated outflowor circumstellar disc. During outburst, hydrogen and he-lium lines appear to have two components, a narrow restcomponent and a moderately blueshifted broad compo-nent. Metallic lines are mainly dominated by the narrowcomponent at all times, although some asymmetry is seenin outburst suggesting that a broad component is present.The square profiles of the Fe ii lines can be explained bythe equatorial outflow model, if viewed pole-on, so the nar-row component is likely associated with this. The broadcomponent becomes weaker and narrower on the decline,and almost disappears in quiescence. This may come frommaterial ejected in the outburst rather than from theequatorial outflow. Forbidden lines fall into two categories;[O i] and [Fe ii] show similar line profiles to Fe ii, but witha lack of low velocity material. These profiles could orig-inate from the low density, upper layers of the equatorialoutflow. Other forbidden lines, [N ii] and [O iii] are nar-rower and likely come from a much more extended region.

The outburst mechanism remains undetermined, al-though the outburst was probably precipitated by the pas-sage of the compact object through the equatorial mate-rial. It is unlikely that X-ray heating of any componentis responsible for the optical outburst. Instead the opticaloutburst is likely associated with the expanding remnantproduced by the X-ray outburst, either through directemission from the remnant or as a result of its interac-tion with the circumstellar material. The spectral shapeof the outburst optical continuum, and the presence ofbroad, blue-shifted emission components, are both consis-tent with predictions for supercritical accretion resulting

in ejection of much of the material (Shakura & Sunyaev1973), and the peak mass transfer rate for an equatorialpassage of the compact object is indeed predicted to bewell above the Eddington limit.

After the outburst changes in the emission lines persistfor at least three years, with Fe ii lines stronger than beforeand He i, He ii, and N ii lines weaker. The timescale for theextended Fe ii decay, at least, is similar to the expectedviscous timescale of the disc of hundreds to thousands ofdays, so this may indicate the gradual recovery of the discto its equilibrium state. As the system does not yet appearto have stabilised continued monitoring is important todetermine if the system eventually recovers to the pre-outburst state or if it settles to a different level.

Acknowledgements. We would like to thank Simon Jeffrey,Amy Mioduszewski, Guy Pooley, John Porter, Rob Robinson,and Lev Titarchuk for information and many helpful thoughtsand discussions which have helped us converge on the picture,albeit incomplete, that we now have of CI Cam. RIH wouldparticularly like to thank Rob Robinson for access to an an-notated high resolution spectrum of CI Cam which proved in-valuable in identifying lines, and for permission to reproducethe data shown in Fig. 12b.

RIH, PAC, and CAH acknowledge support from grantF/00-180/A from the Leverhulme Trust. EAB and SNFacknowledge support from Russian RFBR grant N00-02-16588. MRG acknowledges the support of NASA/LTSA grantNAG5-10889. PR acknowledges support via the EuropeanUnion Training and Mobility of Researchers Network GrantERBFMRX/CT98/0195. WFW was supported in part by theNSF through grant AST-9731416.

The William Herschel Telescope is operated on the is-land of La Palma by the Isaac Newton Group in the SpanishObservatorio del Roque de los Muchachos of the Institutode Astrofısica de Canarias. The G. D. Cassini telescopeis operated at the Loiano Observatory by the OsservatorioAstronomico di Bologna. Skinakas Observatory is a collabo-rative project of the University of Crete, the Foundation forResearch and Technology-Hellas and the Max-Planck-Institutfur Extraterrestrische Physik. This work also uses archival ob-servations made at Observatoire de Haute Provence (CNRS),France. We would like to thank K. Belle, P. Berlind, N. V.Borisov, M. Calkins, A. Marco, D. N. Monin, S. A. Pustilnik,H. Quaintrell, T. A. Sheikina, J. M. Torrejon, A. V. Ugryumov,G. G. Valyavin, and R. M. Wagner for assistance with some ofthe observations.

This work has made use of the NASA Astrophysics DataSystem Abstract Service, the Vienna Atomic Line Database(Kupka et al. 1999) and Peter van Hoof’s Atomic Line Listv2.04 (http://www.pa.uky.edu/∼peter/atomic/).

References

Allen, D. A., Swings, J. P. 1976, A&A, 47, 293

Barsukova, E. A., Fabrika, S. N., Pustilnik, S. A., Ugryumov,A. V. 1998, Bull. Spec. Astrophys. Obs., 45, 147

Barsukova, E. A., Borisov, N. V., Goranskij, V. P., Lyuty, V.M., Metlova, N. V. 2002, Astron. Rep., 46, 275

Bondi, H. 1952, MNRAS, 112, 195

Brand, J., Blitz, L. 1993, A&A, 275, 67


Belloni, T., Dieters, S., van den Ancker, M. E., et al. 1999,ApJ, 527, 345

Bergner, Yu. K., Miroshnichenko, A. S., Yudin, R. V. et al.1995, A&AS, 112, 221

Bjorkman, J. E. 1998, B[e] stars. eds. Jaschek, C. and Hubert,A.M. Kluwer Academic Publishers, pg. 189.

Chakrabarty, D., Roche, P. 1997, ApJ, 489, 254Chkhikvadze, Y. N. 1970, Astrofizika, 6, 65Clark, J. S., Steele, I. A., Fender, R. P., Coe, M. J. 1999, A&A,

348, 888Clark, J. S., Miroshnichenko, A. S., Larionov, V. M., et al.

2000, A&A, 356, 50Downes, R. A. 1984, PASP, 96, 807Drabek, S. V., Kopylov, I. M., Somov, N. N., Somova, T. A.

1986, Bull. Spec. Astrophys. Obs., 22, 64Fabrika, S. N. 1997, Ap&SS, 252, 439Fitzpatrick, E. L. 1999, PASP, 111, 63Frontera, F., Orlandini, M., Amati, L., et al. 1998, A&A, 339,

69.Garcia, M. R., Berlind, P., Barton, E., et al. 1998, IAU Circ.

6865Grandi, S. A. 1980, ApJ, 238, 10Herbig, G. H. 1975, ApJ, 196, 129Herbig, G. H. 1993, ApJ, 407, 142Hjellming, R. M., Mioduszewski, A. J. 1998a, IAU Circ. 6857Hjellming, R. M., Mioduszewski, A. J. 1998b, IAU Circ. 6862Humphreys, R. M. 1978, ApJS, 38, 309Hynes, R. I., Roche, P., Haswell, C. A., et al. 1998, IAU Circ.

6871.Ikeda, Y., Kawabata, K. S., Akitaya, H. 2000, A&A, 355, 256Jaschek, C., Andrillat, Y. 2000, in The Be Phenomenon in

Early-Type Stars, IAU Coll. 175, eds. M. A. Smith, H. F.Henrichs & J. Fabregat, ASP Conf. Ser. 214, ASP

Krelowski, J., Walker, G. A. H., Grieve, G. R., Hill, G. M.1987, ApJ, 316, 449

Kupka, F., Piskunov, N., Ryabchikova, T. A., Stempels, H. C.,Weiss, W. W. 1999, A&AS, 138, 119

Lamers, H. J. G. L. M., Zickgraf, F. -J., de Winter, D.,Houziaux, L., Zorec, J. 1998, A&A, 340, 117

McClintock, J. E., Canizares, C. R., Tarter, C. B. 1975, ApJ,198, 641

Marshall, F. E., Strohmayer, T. E., Lewin, W. H. G., et al.1998, IAU Circ 6857

Merrill, P. W., Burwell, C. G. 1933, ApJ, 78, 87Miroshnichenko, A. S. 1995, Astron. & Astrophys. Trans., 6,

251Miroshnichenko, A. S. 1996, A&A, 312, 941Munari, U., Zwitter, T. 1997, A&A, 318, 269Narayan, R., Garcia, M. R., McClintock, J. E. 2001, to

appear in Proc. IX Marcel Grossmann Meeting, eds.V. Gurzadyan, R. Jantzen, R. Ruffini, World Scientific,Singapore, astro-ph/0107387

Okazaki, A. T. 2001, PASJ, 53, 119Okuda, T., Fujita, M. 2000, PASJ, 52, L5Olling, R. F., Merrifield M. R. 1998, MNRAS, 297, 943Orlandini, M., Parmar, A. N., Frontera, F., et al. 2000, A&A,

356, 163Orr, A., Parmar, A. N., Orlandini M., et al. 1998, A&A, 340,

L19Osterbrock, D. E. 1989, Astrophysics of Gaseous Nebulae

and Active Galactic Nuclei, University Science Books, MillValley, California

Oudmaijer, R. D., Proga, D., Drew, J. E., de Winter, D. 1998,MNRAS, 300, 1700.

Revnivtsev, M., Emelyanov, A., Borozdin, K. 1999, Astro.Lett., 25, 294

Revnivtsev, M., Sunyaev, R., Gilfanov, M., Churazov, E. 2002,A&A, 385, 904

Robinson, E. L., Welsh, W. F., Adams, M. T., Cornell, M.E.1998, IAU Circ. 6862

Robinson, E. L., Ivans, I. I., Welsh, W. F. 2002, ApJ, 565, 1169Shakura, N. I., Sunyaev, R. A. 1973, A&A, 24, 337Smith, D., Remillard, R., Swank, J., Smith, E. 1998, IAU Circ.

6855Stellingwerf, R. F. 1978, ApJ, 224, 953Traub, W. A., Millan-Gabet, R. S., Garcia, M. R. 1998, BAAS,

30, 1330Ueda, Y., Ishida, M., Inoue, H., et al. 1998, ApJ, 508, L167van Paradijs, J., McClintock, J. E. 1994, A&A, 290, 133Wagner, R. M., Starrfield, S. G. 1998, IAU Circ. 6857Walborn, N. R., Parker, J. Wm., Nichols, J. S. 1995, IUE Atlas

of B-type Spectra from 1200 to 1900 A, NASA ref. Pub.1363

Wood, K., Brown, J. C., Fox, G. F. 1993, A&A, 271, 492Zickgraf, F. -J., Wolf, B., Stahl, O., Leitherer, C., Klare, G.

1985, A&A, 143, 421Zickgraf F. -J., Wolf B., Stahl O., Leitherer C., Appenzeller I.

1986, A&A, 163, 119Zorec, J. 1998, in B[e] Stars, eds. A. M. Hubert, C. Jaschek,

Kluwer, Dordrecht, p1

spectroscopic observations of the candidate sgb[e]/x-ray ...2 r. i. hynes et al.: spectroscopic...

Documents